Based on the public and private info I have, here’s how it happened:

It started as an open-source project called gpt-engineer by founder and CEO, Anton Osika: https://github.com/AntonOsika/gpt-engineer .

50K stars, 7K forks, people loved it. But not technical people. Mainly non-technicals.

Anton found the technology fascinating and managed to produce a somewhat working AI agent. ‘Somewhat’ because it was not working every time. He was CTO of a VC-backed company at the time, so this experiment couldn’t be his full time thing.

Anton recruited Fabian as co-founder and CTO, sharing the exciting idea of starting this company together. As far as I remember, on a early morning walk in Stockholm.

The proof of concept was there but the models were not good enough back then.

June 2024, Anthropic dropped Claude 3.5 Sonnet. This was the model, which pushed EVERY AI company forward. This was the breakthrough we were all waiting for. It was really good at coding and could self-heal.

gpt-engineer became 100x more powerful overnight because of this.

Anton was talking to Staffan (Founder of Creandum, one of the leading VCs in Europe, but more importantly- based in Sweden). The funding was largely secured in a very aligned way.

Having worked at multiple startups + a YC alumnus, Anton knew well how to build a VC-backed company:

• Big vision → Trillion-dollar company, solving an extremely big use case (building the last piece of software)

• Innovative product, which relies on technology getting better, so not there yet (awaiting Claude 3.5 Sonnet or similar but ready the meantime)

• Amazing team, working from the same location (in person!)

With all the components there they needed one more thing- to launch the company.

Branding

Lovable’s branding is key to their success.

I can imagine that they came up with the name Lovable and then looked for a domain. Lovable.com is apparently taken. But they’ve gone with Lovable.dev - after all, Lovable allows all people to develop software, so .dev is just right.

They hired a branding agency called Primary, who worked on the branding for just 1 week. That’s how Primary does stuff.

‘Create anything your heart desires’ is hardly the best message. Nor is ‘Your superhuman full stack engineer’ (so lame). Also the AI star in the image is horrible.

But the colors, gradients, feel.

They were all there.

And this was the start.

The Lovable team had been thinking about the slogans quite a lot.

That’s very exicting. Both founders have it in their bio line. Super visionary!

But they have a BUNCH of those.

That’s quite cool, too. Definitely influenced by Paul Graham and YC’s ‘Make something people want’.

They don’t stop there. They speak in super catchy phrases.

Are you kidding me?

This branding and personal branding is definitely strategic. Cocky, infuriating. Also inspiring.

When you deliver on so many levels, you are allowed to do that.

The launch(es)

Here’s a Miro map with the Lovable journey, created by Restartt.

I love how every 3rd one is: ‘launches’ for the first time.

There are like 7 different launches And they’ve been working on it for 2+ years, but the story is: Just 7 months needed to hit glory.

The proper timeline:

• 12 Jun 2023 launched GPT Engineer

• 14 Feb 2024 launched lovable.dev domain

• 28 Aug 2024 GPT Engineer launches out of beta

• 7 Oct 2024 Lovable raises $7.5M with ‘no product’

• 10 Oct 2024 Lovable reaches 1200 paying users ‘before launching’

• 19 Nov 2024 Anton Osika says GPT Engineer is now Lovable

• 21 Nov 2024 Lovable officially launches

Lovable launched at least 5 times. If you ask me- 15 times.

And that’s how it should be.

Hype after hype after hype.

At the end of the day, AI products are changing so fast, that if you have tried a product 3 months ago… you haven’t really tried it.

I talk about this here: AI Agents Are Here: The Adoption Methodology.

Also Anton on this:

The team

Lovable’s team is absolutely cracked.

They hire only the best of the best.

Preference? Founders.

To quote Anton:

This is why we systematically hire people who can build something from nothing. Very often previous founders, who have run their own company and know what it feels like to make hard calls with incomplete information. To feel the weight of it and keep going. That instinct is hard to teach.

Often these are ex-YC founders.

People who care about equite, responsibility, ownership, impact.

Enzo Avigo (co-founder of June.so, exited to Amplitude) was in talks with Lovable as well. After spending a week together, Lovable wanted to only hire Ferruccio, his co-founder (CTO). And Enzo is fucking top notch, too. Getting both Enzo and Ferru on my team sounds awesome and would take them any day of the week. Not at Lovable.

No compromise with team.

And that’s how it should be.

They have 18y old geniuses. They have 40+ old veterans. Their Growth team is led non other than Elena Verna.

People from all ages are drawn into this. It’s a generation-defining company.

Lovable is the benchmark for all AI startups.

Conclusion

I’m sure I’ve missed context that only insiders would know, and I don’t want to mythologize anyone’s work.

But even from the outside, this looks less like a miracle and more like a series of high-quality decisions made fast and then repeated.

Lovable’s story is worth studying in detail.

August 2024 was the best month of my life.

I closed the biggest deal of my life: $4.5M in Seed funding for Team-GPT [now Juma].

But what made it special was not the achievement itself.

It was two other things:

1. The struggle I went through to get there

2. The fact that I shared the achievement with my loved ones

The struggle

Starting Team-GPT was not all rainbows and butterflies.

We had a lot of internal fighting since Day 1. The initial equity distribution was a total mess. It was unfair and would not yield a prosperous company.

Reaching a lean and fair distribution of the equity took us 6 months, and another 4 months to get it done on paper. This was a major risk for the project itself and caused a lot of anxiety and trouble to everyone involved.

But it had to be done for the good of the company.

Not only this.

On the 3 Jan 24, my dog died. It happened very fast. I was in Austria for New Year’s Eve and drove 1000km in 1 day to see him for the last time, but I missed him. It broke my heart twice as much.

It was the worst start of the year.

My mom said:

When you go fundraising, you’ll forget about the pain.

I went to fundraise for Team-GPT shortly after.

I was looking forward to this challenge.

It was something I had never done in my life.

I knew it was going to help me grow as a person. I needed to go through adversity to come out stronger. I went 100% out of my comfort zone.

Everyone I was talking to was telling me I’m asking for too much money, no one is going to give it to us.

I didn’t care. I knew how much I needed to make my vision come true.

But people doubting me so much was a bit new for me.

It was getting to me. It was shaking me to the core. I was starting to doubt myself and the business.

The rejection was killing me.

So, there I was. All alone in San Francisco- home of all AI and VC funding.

I was almost destroyed. A complete reck.

My sleep score was at an all-time low.

Stress – at an all-time high.

I came home for Easter.

I couldn’t do it any longer.

After 2 months abroad, the fundraise was not done yet. This meant the company would seize to exist and I would have let down everybody involved.

Came home to my family to celebrate.

They were hoping for me to talk about the whole experience. However, all they got was an Iliya who was not speaking for days.

I had been pitching for 15 hours a day, every day. For 2 months.

I didn’t know what to say.

I couldn’t look my father in the eye.

I was trying to smile but it was not coming out well.

They asked me whether I’ll quit trying.

I had made the mistake of joking that ‘I’d rather kill myself than stop in the middle of it’. This really concerned them about my mental health.

And indeed, I had a full pipeline of VC calls. I knew I was very close to raising money.

People wanted to invest!

I was just too picky – I wanted to find the right partner.

The intervention

There was a family intervention. They thought I was suicidal. It was unacceptable for my family to keep looking at my suffering.

I looked at them and said:

You don’t know what I know. I have been in these meetings. I know how carefully VCs listen to me. I know we are onto something. This could be the project of my life. I’ve prepared perfectly well and I’ll deliver!

A family member said:

I have no doubt that you’ve prepared better than everyone else. But we have to face the reality at some point.

I was absolutely crushed.

20 days later I had received 2 term sheets, essentially filling the $4.5M Seed round.

I was ready to take everything above $3.5M, so this was a major win.

I came home and took a 1-week vacation with my family.

It was raining.

Celebrated my birthday with Prosecco in the morning.

Listened to classical music all day: Bach, Beethoven, Mozart, Schubert, Brahms, Haydn, Wagner, Tchaikovsky, Chopin, Stravinsky, Verdi.

We did them all.

My brainwaves were coming back to normal.

Especially because I was back with my family.

Best month of my life

So why was August 2024 the best month of my life?

I was deeply struggling between January and May.

June was the month I came back to life.

July was all about getting the documents ready and hiring new key people to the team.

August was the month where everything fell in place. August is when it all unraveled.

But mainly → the fact that I shared the achievement with my loved ones

I was back with my family.

But I was also back with my team.

I fuc*king love our team.

I’m so energized when I’m around these people.

In Aug 2024 we had our first major offsite.

We called ‘Summer Brain Camp’

We took the time to plan the next few months.

We bonded.

We were together when the lawyers sent us the documents for the financing.

We were together when we signed.

We were together when we were relieved from all the anxiety.

We were together when we shared the dream of becoming a unicorn.

Several days later, all was closed.

We waited for the money in the bank for 7 days.

Every day we were checking together, hoping we will be together in one place when they arrive.

On my last day on the seaside (I was born in Varna, Bulgaria – where my parents live), just as I was about to leave, we decided to check if the money was in the bank one last time.

It was a magical moment.

We had done it.

And we could share the moment together! ❤️

It’s hard to explain how happy I felt at the end of it.

Truly the happiest time of my life.

To savor the triumphs, you absolutely need the struggles.

Make sure you share both with your loved ones – it provides all the meaning.

See also:

The 100% real story about our fundraising. Journalled as it was happening.

👉 How Team-GPT raised $4.5M- A True story

Note: Throughout this article AI refers mainly to GenAI, e.g. ChatGPT, Claude and agents like Juma, Manus, Lovable.

I’ve been thinking a lot about the best path to AI adoption.

How do you learn it? How do you teach it? What does it mean to adopt ChatGPT well? What is stopping teams from adopting GPT? Where is the resistance coming from? Why are people achieving such wildly different results? Why does MIT state that 95% of all AI adoption projects fail?

Most humans will be left behind

Since ChatGPT came out the world really changed. This technological leap is historically unmatched. The world economy is majorly, irreversably, and inevitably changing and technologically advanced nations and individuals will have much better outcomes in life on average.

The problem is that many people who I respect are denying or soft-denying (not using) AI. Maybe 1 in 5 of my friends use it passably well. At the same time, people at our company (Juma) have mastered AI to a point at which, they’ve become 10x to 100x more productive than non-AI enabled humans.

The divide is becoming bigger and bigger by the day and I don’t want my friends left behind. In fact, I don’t want anyone to be left behind.

But the reality is that 90-95% of human population will be left behind in some form. My approximation is that 2-5 billion people will have to adopt AI in the next 10-15 years but only 200-300 million of them + AI will be doing most of the digital work.

It should be anyone’s goal to be one of those who have a job, once the AI industry matures.

People don’t understand why this AI is so different and powerful

‘Smart people’ are saying AI is the future. But even they don’t know why…

So, why?

Partly it’s because the whole Internet is condensed into a single model that can run on your iPhone without Internet connection.

But the more prominent and more urgent reason is: AI agents.

Not any agents but coding agents in particular.

It’s becoming more and more clear that:

• conversation (chat) is the universal user experience to communicate with AI

• code is the universal interface to get work done well (watch this by Anthropic)

You can dictate to an app and it can write code to create any software… on the fly. We’ve already seen the massive success of coding agents like Lovable and Cursor.

Since most of the work use cases can be solved by code, this is just the beginning of coding agents and their dominance.

Everything automatable will be automated.

But humans will still be the ones giving tasks and objectives to the AI… Right?

How to adopt AI for work?

Obviously, creating a ChatGPT account is not enough. People need to actually use it and explore its bounderies every day.

Many people create an account, stay on the free tier, and ask basic questions which Google is perfectly capable of answering.

AI can search the internet but also create images, presentations, strategize, work with documents (Word, Excel, PDF) and analyze data.

But we haven’t seen anything yet. At the time of writing, we are only 3 years into ‘post-ChatGPT’ era and I believe that any decent developer can build 99+% of the software ever built… by prompting an AI agent. Everything is possible and as easy as explaining the problem you want solved.

The capabilities of AI are growing daily. Knowing how to use AI is a moving target.

• Any webinar discusses past findings

• Any workshop teaches the best practices of yesterday

• Any how-to video is old the moment it’s published

Start by lighting the spark

We, teachers, have a saying:

Often our job is to simply light the spark in a student. It’s up to them to put fuel in the fire.

As much as we want our friends, students, colleagues, compatriots to jump into adopting AI, we must begin by lighting a spark.

They need to get excited!

WOW them somehow.

e.g. get 4h of work done with a single prompt

Ensure you’ve WOWed each person you’ll be helping adopt AI.

Educating the workforce about AI

Building knowledge and skills is hierarchical*.

Effective learning is about building a fundament and then stacking layers of knowledge on top of it. All else equal, the stronger the fundament, the better the outcomes.

For example, school is the fundament of everything (for people who were schooled). If you have never been to school, everything else is very hard. You build on top of school with university, then professional experience, etc. The more you stack, the higher you go up the pyramid, the more masterful you become. To adopt AI and to learn to utilize GPT-like models, we need to build a strong AI fundament.

The Pyramid of Human+AI Knowledge is a useful helper to explain The AI Adoption Methodology.

There are 3 layers:

• Fundament
  Build a strong fundament and it will hold the pyramid.

• Context
  Provide the relevant context to the AI- your coding repository, your Google Drive, your Sharepoint, meeting recordings, or anything else. Make it specific and relevant to your organization.

• Mastery
  This is when AI is seamlessly integrated everything you do

Full AI Adoption must be = Mastery.

The Fundament

You want your people to know how to use AI. Not how to train AI models.

Too late to train AI models

If you have not been in AI for the past many years, forget about building a career creating AI. You will be harnessing other people’s AI at best. And that’s alright.

It’s too late to train AI models. I have been teaching people how to train these models. That was almost 10 years ago…

The people who are required to train AI models worldwide are probably < 10,000 overall. We can argue that 10-100 people are really pushing the technology forward.

Forget about training AI, optimize for using it well.

The ‘Aha’ moment + knowing what AI can do and then prompting it well

Initially, our hypothesis was that there is a critical number of chats before you reach the ‘aha’ moment. The moment when you simply know that GPT is something else entirely. My guesstimate (based on the people I’ve helped learn) is that to ‘get AI’ you need 3-5-8 conversations on average, depending on the person and their own AI inclination. I’m convinced that at the 10^th fruitful conversation, even the biggest sceptic shall be converted to an AI fan. This became the premise of the ‘ChatGPT for Work: The Interactive Course’, where we had cherry-picked 10 convos that are worth having + 100s of exercises.

Today, I think that 1 chat is enough to reach the ‘aha’ moment. Probably 500+ million people worldwide have had a chat like this. Yet 99%+ of them have barely scratched ths surface of what’s possible with AI.

We made an attempt to teach people how to use ChatGPT.

Althought somewhat successful, 2 years after ‘ChatGPT for Work: The Interactive Course’, I think we should rethink education fundamentally.

The AI industry will evolve in the folowing way.

AI Agents will:

• Have wide variety of expertise

• Be mostly a chat interface, where you prompt or dictate

• Behave like an agent (instead of assistant/chatbot), not only answering questions but also executing series of tasks. Sometimes for hours or days on end.

• Have access to many tools (in our case that’s marketing specific tools, e.g. keyword research tool, post analysis tool, Google/Meta ad analysis tool, etc.)

• Access to many hidden assistants (pre-made prompts that are tried and tested- best practices), so that the user doesn’t need to know them

• Have a single model, which uses different models under the hood, depending on the task

These products will be guiding the user into what’s possible.

And then, they’ll get it done.

You won’t need to know how to do something, as long as you know what is possible today and if this particular AI can get it done.

Core concepts which probably won’t change

There are only a handful of good practices that are evergreen.

1. Foundational Concepts

   1. AI works on probability and pattern recognition, not true understanding—this limitation shapes how to use it effectively

   2. That providing rich context is more valuable than clever prompt engineering tricks

   3. Leveraging multimodal inputs (screenshots, images, documents) as context for more effective prompting

2. Core prompting fundamentals- Providing clear task and context about it, youeself, etc.), asking AI clarifying questions before diving into solutions, rather than trying to craft perfect prompts from scratch

3. Universal AI interaction patterns

   1. Working iteratively with AI through conversation rather than expecting perfect first-draft results

   2. Using personas/roles to quickly establish context and expertise domains with minimal text

   3. Prompting by example (scientifically most useful technique)

Here’s one evergreen document from ‘ChatGPT for Work: The Interactive Course’.

Prompting Pdf

3.73MB ∙ PDF file

Download

Download

Everything else changes daily.

The products (agents) are becoming smarter by the day.

It’s an open question which product can do what. Also, just 3 months after you try an agent, it is probably 10x better.

Agents improve rapidly, so interactive education is needed where the AI decides what skill to teach

As an attempt to be futureproof, we made Juma Journey - an interactive learning experience, consisting of 10 practical missions, showcasing different capabilities of the Juma superagent.

The course doesn’t show stale information. You constantly communicate with the AI model itself, asking it what it can do, storing it in its memory and progressing further. This is a self-improving online course. As Juma becomes more powerful, the course fundamentally changes.

5 years from now, the same course will help people achieve completely different goals.

I don’t know what will be possible. I just know the model will guide users.

The Context Layer

Once you’ve placed the Fundament, i.e. understood why AI is the next big thing and how you can prompt it, it’s time to make it actually useful.

Even if you have the best ‘prompt engineer’ on your team, they are only as good as the data or Context they provide to the AI.

A good prompt engineer is a good domain expert.

For instance, I’m awesome at prompting- probably better than almost any lawyer on Earth! But almost any lawyer (who knows the fundamentals) will be a much better law/legal prompt engineer than me. Why? Because they can define the tasks to be done and can provide better context about law than me.

Domain expertise matters a lot because only experts know what context is important (and what isn’t).

To make AI work for you, you need domain experts to use it well. These will be providing the right context, at the right time.

1. This can happen naturally (in the span of a chat), in which case you’d be set up for failure because:

   1. you depend too much on the memory of the particular product and

   2. this is an individual (and not collaborative experience).

2. Or it can be a focused effort, where 1 or more people organize the info in a collaborative Project or similar, e.g. Juma Projects.

Some professions have it easier, e.g.

• Developers work in repositories, which they unify through GitHub for instance. So it’s already super organized: you connect the repo and it has all the Context for the project at hand.

• Marketers work on campaigns, which often have a start and end date and must be executed rather fast, context is created at the start and evolves for each campaign

• Agencies collaborate on clients, ideally in the span of years, so they often have this stored somehow

More often than not, context is scattered across tools, drives, archives.

An absolute key step to adopting AI well for your organization is to cherry-pick the right context that the AI will be stepping on.

This is so important, that nowadays we often talk about Context Engineering as opposed to Prompt Engineering.

Context is King.

Cherry-picking the right Context

In 2026, Context for coding is pretty much sorted. For everything else, we need to make an organized effort to collect the right data.

For instance, as I’m writing this, an AI agent is analyzing 1 year of meetings (a total of 1500 meetings) looking to identify pain points, common objections, winning phrases across sales calls. The agent will be running for ~3 hours and then I’ll be analyzing this data in Juma. There are also other agents on the market that can solve this problem.

But none of them can solve it without me providing the 1500 meeting recordings/transcripts.

Context is important and providing it in the easiest form possible to the agent is a key skill.

The right Context at the right time

You can’t expect an AI to always know which context to use when.

Just think about your own knowledge and memories. The fresher they are, the more important they are. Thoughts, ideas, memories from 10 years ago should often be disregarded or at least discounted.

Now think about your company documents. Documents that are 10-20-30 years must NOT occupy the same space as documents from this year. But sometimes they do.

Often the people on your team don’t know what’s important and what’s not.

How do you expect the AI agent to work it out?

Using the right context at the right time can largely be solved by agents one day. But I cannot imagine it solving it completely.

Humans will curate Context for a long time

I think for much longer than expected, humans will be curating the Context for the AI.

Whether it’s PDFs, images, integrations, audio files, videos, social media accounts, web pages, MCP servers, or whatever is trendy in the future… it will be us making sure the right context is provided and/or used.

As we create the products of the future, we fantasize about the AI being able to select the right context at the right time, but unless we transmit our thoughts directly via Neuralink, I hardly see it happening.

In-house AI Champions

Internal AI champions are the backbone of the AI revolution.

All organizations that have successfully adopted AI have 1 thing in common:

They treat adoption as a project with a clear owner, who is responsible for getting it done.

Who are these AI champions

These are people who are usually geniunely excited about the technology and want to make their collegues successful. They champion AI because they see it as an opportunity to fast-track their career.

Ideally, you want to find someone who hits both of these.

They will likely be young(er), tech-savvy, and rather ambitious.

You should give them full responsibility (and accountability), send them this resource, and empower them as much as possible.

What does an AI champion do?

There are many jobs that can be done, here’s a non-exhaustive list: know their team and their AI aptitude, make sure everyone has access to AI, make sure everyone uses the company AI, define KPIs and later measure the success of the rollout, organize workshops, 1:1s, educational sessions, find the use cases that matter and surface them.

Here’s an example 12-week AI adoption plan to start from.

Ai Adoption Plan 12 Week

7.57KB ∙ XLSX file

Download

Download

Ai Adoption Plan 12 Week

108KB ∙ PDF file

Download

Download

I would advise against hiring external people to get this done. You want it to be someone who knows how the company works from the inside.

Finally, AI champions must work individually with every person to make sure they see direct benefit of AI to their work. If that’s not feasible, they must find ‘sub-champions’ within each department.

AI champions and Context

The first job of the AI champion is to ensure everyone in the org has the Fundament in place.

The second (and most important job) is to to bring the domain expertise of the company inside the software in question.

Mastery

At the top of the pyramid is Human+AI Mastery.

Mastery is when it becomes hard to distinguish if something was done by human or AI.

Mastery is when you stop thinking about ‘how can the AI help me’ and just know it.

Mastery is when you dream bigger because the AI is already an extension of yourself.

I really believe the right metaphor is: extension of one’s self.

AI as an extension of the human brain

When the first humans started using tools: hammerstones, stone cores, sharp stone flakes, this allowed them to hunt, build houses and so on. Without going through the whole history of humanity, tools have been very useful and were practically the big reason why humanity was evolving.

When talking about tools, there is a very interesting neuro-phenomenon.

I’ll quote MIT Press on this:

When a tool in your hand “becomes part of you,” it’s not just a metaphor… It’s real. Your brain makes it real. Even if the brain just pretends that the tool is part of its body, then the tool is a part of its body.

For example, when you are inside a car, especially driving, you become one with the car. You stop thinking about the car, you are the car. In every maneuver in your head, you are no longer just human. You are also a car.

In the same way, pilots are one with the plane, captains are one with the ship. You are experiencing a similar feeling when holding: a tennis racket, baseball bat, knife, phone, computer mouse, or any other object that can be used as a tool. You are extended with these tools.

This is true Mastery. It’s not just not just a metaphor.

It’s real because your brain makes it real.

Can you afford to skip on AI adoption?

The times they are a-changin’

10 years from now, every company will be an AI company.

People who have adopted AI already have an unfair advantage. They are not only more productive but also disproportionately more capable.

Much bigger goals are now conceived and realized by smaller groups of people as it takes less time and resources to get things done.

For those behind- this is your last call to get serious about AI adoption.

For those ahead- technology is accelerating at an ever-faster rate. Whole businesses are erased after a model upgrade. In this new economy, nothing is defensible. Not even your 3-year AI adoption advantange.

• Juma Journey - 30-min AI learning experience to build the Fundament

• Juma - marketing superagent

• Connect with Iliya on LinkedIn

Special moments don’t just happen.

More often than not – they are created.

Carefully crafted by someone or a maybe a whole group of people.

Think about an obvious one- a marriage proposal. Every marriage proposal is thought through by one of the sides, often an engagement ring is chosen, the words are rehearsed before the actual proposal happens.

In a similar way, to appreciate the special moments in life, they need to be created.

Created. Crafted. Curated.

Celebrating a birthday is about creating an event, curating the people who will attend, choosing the venue, music, etc.

Some special moments are as simple as asking a friend for a walk to tell them how much you appreciate them.

The thing is:

• You can hardly create a moment over a phone call

• You can’t be vulnerable in a bus full of people

• You can’t be fully yourself on social media

Legendary moments are carefully crafted.

I’m not saying planned to the detail or rehearsed till the point no emotion is left.

I’m saying-
A safe sentimental environment is needed for good moments to be shared, remembered and cherished.

The Moment that gave us wings

The moment we received a term sheet from True Ventures was such.

We were invited to New York to meet the True team.

We got a verbal offer but negotiated the terms for 20+ hours.

In the end, I was invited to an event, in the middle of which, in a back room, we got to talk for ~30 mins.

This is when we received a signed paper copy of the terms, that were finalized just 1 hour before.

5+ people from True were involved to get it done on that day- in person, at the very center of New York city.

We worked so hard to get to this moment.

Truly lifechanging for our whole team.

But such moments don’t just happen.

They are created.

Thankful to the True team for everything they’ve done for us.

Subscribe now

I’ve gone from $0 to $1M/year for the 3rd time in my life.

Here are 25 startup truths:

1. Building a great team takes time. I’m building ours for 9 years, across projects.

2. Radical transparency builds trust like nothing else. Mostly- attracts the right people, repels the wrong ones.

3. It’s cheap to build software, it’s extremely expensive to build a great product.

4. Bootstrapping vs being venture-backed depends on the market and competition. Not on your personal preference.

5. PR is vanity. Yet it’s an important part of the marketing mix and must be done.

6. Each time you get from 0 to1 is different. The fact that you’ve done it before helps... but the strategy and tactics are different each time.

7. The true test is: how much $ will you be making if tomorrow you have 0 employees. What about in 1 year?

8. It’s all marketing. Nobody loves their creation on day 1. Yet you can’t make $1M if you don’t sell the s**t out of whatever you have.

9. The more senior you become, the more sales you do.

10. CEO needs to have deep expertise in the field of the business.

11. If you want to go fast, go alone. If you want to go far, join forces with others.

12. Owning 5% of a $1,000,000,000 business is better than owning 95% of nothing.

13. 30% of the time of a CEO goes into admin and finance. As CEO, find someone to own this ASAP and you’ll accelerate the trajectory of your business.

14. If someone is not working out, part ways fast. It’s better for everyone.

15. Toxic clients should be fired.

16. Partners that benefit from your partnership at your expense should be removed. It must be a win-win.

17. Look for ways to outsource as much as possible. Never outsource your core product/activity.

18. Always reward hard work and dedication. If you can’t afford to do it with $$$, give more time, attention, coaching. That’s often more valuable.

19. MacBook is orders of magnitude better than any Windows computer. You can’t reach your full potential without a Mac. Not an opinion but a fact!

20. Great systems are the difference between very good and truly great.

21. Process is the opposite of creativity. To stumble upon great ideas, you must avoid putting people in boxes. Yet no process means no scale.

22. Your 20s must be spent in experimentation. Find what you hate. Find what makes you tick.

23. It’s easy to give back. It’s hard to pay forward.

24. Being direct may be perceived as being rude. Yet always being direct makes you predictable and trustworthy, which is extremely productive.

25. It has never been easier to start a business. Yet it’s harder than ever to build lasting value.

This post was based on my experiences scaling businesses to $1M.

Yet money has never been the final goal for me.

How can the means of getting there be the destination?

Hey!

Our best month ever in terms of ARR, just shy of our most important milestone ever. Depending on how we count it and when you are reading this, we’ve hit $1M ARR. Yay!

Yet we don’t feel like a million bucks.

We’ve got so much potential and somehow we can’t seem to properly unlock it. Yet.

As of 1 Oct 2025:

• ARR: ~$969,288 (+$137,155 vs last month, 16% MoM growth)

• # Users: 67,606 (+7,264 vs last month)

• Burn rate (approx): $145,424.05

• Runway: 17 months

• # People: 20 full time + 3 part-time (+ 1 → Simona is back!)

I. Built for everyone, optimized for no one

I think the biggest issue of Team-GPT from Day 1 has been that we sell to everybody, or better phrased: we don’t know who we sell best to.

This has undermined every decision we’ve made (because we never commit fully).

• It’s the reason why some investors couldn’t build conviction around our vision.

• It’s the reason why clients don’t buy our product (they don’t understand how we are useful to them).

• It’s also why people say we are ‘just a GPT wrapper’, even though we’ve got lots of IP on top the LLMs.

We’ve been brutally aware that we are ‘built for everyone, optimized for no one’.

But as CEO, I never had conviction in the right focus for the company to just say ‘GO’.

I only had conviction in our team’s ability to make awesome things happen… fast.

Btw I’ve always been upfront about it: ‘We need 30-50 brilliant people working together to build an incredibly impactful B2B AI product’. N.B. We are at 24.

Together we’ve made lots of steps towards understanding product-market fit (PMF) and where it will happen for us. We’ve invested more than almost any other startup at our stage in analyzing our data and understanding it well.

II. Mission should never change

Jon Callaghan (co-founder of True Ventures ) said:

‘At Seed stage, your company and product will look nothing like it will at year 10. The only thing which will never change is your Mission.’

Our Mission never changed:

We are on a mission to help people adopt AI for their work.

nor the Vision*:

We are doing this by building a future-of-work B2B AI product, which puts humans at the center and promotes collaboration, privacy, and agentic workflows.

*the exact phrasing is work-in-progress.

III. Optimized for learning

To find the fastest route to fulfilling our mission, we’ve taken an extremely data-driven approach.

Our process:

• We optimize for learning

• The product is the playground where we experiment

• Extreme PMF + global impact is the North star

• $$$ value (ARR) and usage metrics are the ‘weights and biases’

It’s been almost 2.5 years since we started ‘learning’ and the results are becoming clearer and clearer.

IV. 1-million-dollar-weighted insights

I think reaching $1M ARR is the key here.

Not because it matters in the grand scheme of things. We are still making ‘peanuts’ and we still can’t support our 24 (brilliant) people without external investment (for which we are eternally grateful to our investors). If we were bootstrapping, as of today, we would be able to afford ~10 people full-time.

But $1M ARR is the key because we’ve gone from 0 to 1:

• We’ve built something from nothing

• We are the go-to software for a small group of people

• If we stop existing today, thousands of people will actually suffer

We matter!

~1,000,000 USD per year are paid to our company bank account because of what we’ve built.

These dollars weigh.

They weigh on the data when we make decisions.

When analyzing our data through the lens of a million bucks, we’ve learned that:

1. Team-GPT main value prop -> Projects (aka assistants/agents)

First and foremost, our main value proportion are the Projects (60% of happy, dollar-weighted users use them).

Projects are own main organization unit + Context/knowledge unit.

Why? Because they solve the main issue with AI → explaining the same context again and again. This includes:

• Project Knowledge - Upload docs, connect Google Drive/SharePoint

• Context Library - Reusable context across the

• Assistants (knowledge bots) - Project / department-specific assistants (HR, Sales, Legal).

• Some users use Projects as folders, most use them like Assistants. From this month they can use them as Agents (see ‘automations’)

2. Multiple models (Claude, GPT, Gemini, etc.).

Biggest value prop in sales calls. Head-to-head with ChatGPT and Claude, this is our only benefit, which becomes clear straight away. In the context of our main value prop (projects), this means: “Before users underhand they actually care about Projects, they care about model availability and no vendor lock-in.”

3. Enterprise Control & Security

Our 2 biggest contracts ($50K+ ARR each) have been signed because of this. We’ve got many open conversations about it and it seems to be a place where we can position ourselves long-term.

These users, however, will NOT help us build a top notch product TODAY.

I think that’s the real reason why successful products like Slack, Notion, Miro, etc. move to enterprise last. We are forced to do the same. But we will try to close all deals that come our way before that.

P.S. These users care about:

• Private instances

• Compliance (SOC2, HIPPA, ISO)

• Zero data retention options

• Permissions and governance functionalities

4., 5., 6. And the rest are all secondary

We believe the combination of our other benefits is valuable. Yet, $-weighed, everything else about our product is much less significant than the above:

• Pages → 9-20% of users that matter (depending on whether you care about $ weight, ICP weight, activation weight, etc.)

• Prompt library → ~15% of users care

• Tools → nobody cares (with the notable exception of Prompt builder)

• Integrations → valuable but not enough (because we don’t have many integrations yet). We understand that to build a successful B2B product, we will need to connect to all other major software

We kind of know what features and benefits matter.

But who cares?

V. Approaching Marketers (our ICP) with conviction

Marketers care.

Experimentation + data showed us that we should narrow down to the marketing function.

We did this in June 2025.

Some called it a ‘pivot’. In reality we simply changed our messaging to marketing with 0 changes to the product. We showed commitment on the outside to validate if this is the right decision to make.

We’ve started making changes to the product but not by making it more flashy to marketers.

We are making the product more useful to our top ICP users, which in turn will make it more appealing to the right marketers.

While we are seeing growth along all categories, marketing seems to be our ‘bread and butter’.

Marketers care.

Looking at the data it is somewhat obvious, isn’t it?

We should approach Marketing ICP with conviction.

VI. On the brink of $1M ARR

We know who cares about us.

We know why they care about us.

But the name Team-GPT got us from 0 to 1… is now holding us back.

People pull out of deals because we are the ‘fake ChatGPT’. With this name we will always be in the shadow of OpenAI.

Worst of all… we don’t have our own identity.

Team-GPT’s time is up.

Long-live XXXX.

New name redacted, will be revealed in November 25.

⚠️ Name is a secret, please don’t spread yet ⚠️

We are rebranding in November and will launch XXXX in San Francisco.

We will build our new identity, which will be fully focused on marketing as ICP.

Now that we know who our ICP is, we will listen much more carefully to what they have to say.

VII. What brought you from 0 to 1, won’t bring you from 1 to 10

We’ve achieved what 96% of startups never will.

Yet I’m far from satisfied, because we’ve got much more to show.

Hard work is just beginning.

This time on a different scale.

I’m thrilled to embark on the journey to $10M ARR!

Time to unlock our true potential ⛽️

Best,

Iliya

The Bridge V was nothing short of extraordinary.

Here’s a summary from my point of view:

Day 1

• Woke up at 8.30AM. Since we were in Varna, my hometown, my mom made me breakfast and told me my clothes were inappropriate. Felt like a teenager again.

• Went to The Bridge and got recognized from afar because of my Gen-Z-emo-girl outfit. Personal branding win + showed mom I was right (or did I?)

• Event started at 11AM with awesome jokes by Olya Antonova and some crazy pep talks by Nikola Likov, main guy behind The Bridge

• Got goosebumps by the the coolest intro video by Strahil Jordanov. Such a cool branding!!!

• Started thinkin’ about an exit, thanks to Jeko Kolev‘s tips and tricks on selling a business. Also, heard about Jeko’s extraordinary life. Trust me, I’ve heard it 1 gazilion times. It was refreshing he can do it in 3 mins.

• Geri Kirilova told the truths about Series A fundraising! She knows what she is talking about!

• Had lunch with Teodosy Teodosiev, early investor in Team-GPT, also my biggest critic. Spoke about electrifying old Porsches. Agreed that once Team-GPT reaches $1.4B valuation, I’ll earn his respect

• Hosted an AI session, where the people had combined net worth of $1B+, e.g. Vassil Terziev (founder of Telerik, one of the biggest BG tech exits + Mayor of Sofia), Konstantin Dzhengozov (co-founder of Payhawk, and since last month, top recruiter for our team), and Dimitar Dimitrov, founder of Ardes.bg, and one of the top education benefactors I know.

• Gave the guys an AI test. Don’t worry- they knew their stuff. Looks like they are above average.

• Moderated a session on ‘Designing life on a new planet’, where Irina Obushtarova looked like our most likely suspect for a spiritual leader, while Kremena Dervenkova promised, she’ll take care of nutrition on the new planet, spoiler alert, we will eat only insects- forever. Konstantin Bezuhanov will lead our electronic identification + all arts and crafts. Here are the notes:

• Had dinner with Alexander Karolev, Kaloyan Popov, Nedko Kolev, and Boris Pavlov -> some of my good friends outside The Bridge

• Walked to the club with Alexandra Todorova - the operational backbone of The Bridge. Absolute machine! Don’t know how but she did it again!

• Celebrated my 31st b-day with everyone above and some more. Didn’t pay for anything. It was people like Angel H. Angelov, CF (one of the few BG investors in OpenAI) who paid my bills.

• Speaking of unicorns and payments... I’m thankful to Kosio for helping with our recruitment efforts at Team-GPT 🩶 We’ll never need an HR with support like yours!

• Discussed life after marriage and kids with Bozhidar Georgiev, who has the best co-working place in Sofia - Work&Share. Yes, I’m biased cause Team-GPT is there, but rest assured that we cannot settle for anything but the best. So it is the best.

• Enjoyed some drinks with the most stylish 4 chaps: Theo S. Shikov, Dimitar Mitkov, Boris Georgiev, and Zdravko Zdravkov. Discussed how their angel investment in Team-GPT is going. They gave me 1000 tips how to sell better and practically sent me to get to work.

• Went to an undisclosed club, where Danail Koev, Ilian Kodzhahristov, and Botty Dimanov, PhD among others absolutely destroyed the s*it out of me.

• Someone broke a glass in my head. Still hurts.

Day 2

• Next day I went back to the Bridge and first person who saw me ( Tony Aleksandrov ) said: ‘You’re were a frickin’ firecracker’. That’s the point at which I knew something might have gone too far.

• Watched and listened to Ivet Lalova‘s incredible story. Seriously, this was probably the highlight of the whole event.

• Had lunch with Victor Botev, Yavor Belakov, and Daniel Rankov. We’ve got many things in common, mainly Several Clouds doing our AWS infra. Approved a PR, where the work Dani’s company had been doing for months would be finally moved forward.

• Went to my first ever Stress Management group. Thanks Galena Tsankova that’s exactly what I needed.

• Picked Boris Valkov‘s brain on how to reach PMF. It was Boris who was the first to tell me how to fundraise, so I know I should listen to him!

• Tried out the isleep matraces by Matrax (and promoted by Ivet). Entered a raffle to win a matrace lol

• Tried to get some merch from the event. There was a white T-shirt. Turned out it was only if you enter a marathon. Had to enter the marathon.

• Prepped for it with pro runners like the Founders of LAUNCHub Ventures, some of their founders, and even their LPs.

• I kind of walked the marathon (2.5km only) because the doctor told me to not run... It was like a walk in the park. Interestingly, I wasn’t last. I beat a pregnant woman to the finish line 🤰

• Remember Ilian Kodzhahristov? The guy who drank in the club until morning? He killed it at the marathon. By a landslide!

• What about the raffle for the matrace? Turns out I won it!!! Best day ever!

• Bought Kalina Panayotova‘s newest book Nectar - looking forward to reading it!

• Went to dinner. Had some awesome talks with Stoyan Ivanov about enterprise AI agents.

• Learned there was an all-white dresscode, which I terribly, terribly missed. Turns out when you are a CEO, you really don’t read 70-90% of the comms that come your way, especially attached in PDFs or similar. Oops.

• After my teenage girl outfit from Day 1, people were probably expecting of me to be great looking woman young, dressed in all-white

• Instead I was a 14y-old off-white teenage boy. At least not a girl, mom!!!

• Discussed with Simeon Simeonov and ⚡ Orlin RadEV ⚡ how Future Unicorns Accelerator has a 100 NPS score, is the best thing ever, and more people should know about it! Also touched upon rebranding it to ‘a bootcamp’ rathern than an accelerator, as it is closer to its purpose. Soon after, things escalated and we discussed the idea to build our own club, since the music in clubs was too loud for Sim.

• Went to the club, nonetheless, and had some more fun. Got told to s**k some d**k, which was refreshingly funny. At least they didn’t break another glass in my head.

• Came home early (4am), just to vibe-code a little side-project (QR code generator) I’ve been meaning to for quite some time: Beautiful QR Codes

• Went to bed happy with the fact that I even deployed it in Vercel. It’s also mobile friendly and 100% free. Thank me later!

Day 3

• Missed the 3rd day of The Bridge.

• Had lunch with mom. Ate some chicken fingers. At this point, she told me I’m definitely a child (gender was not specified this time). But this is how we democrats like it- gender fluid. And prices in EUR.

• Went to the airport, where I met 100 Bridgers. Debriefed with Lyuben Belov - one of my favorite people to bounce ideas off

• Came home to Sofia

• Went out to the park to have some beers with my lovely wife Maria Valchanova in the city center, as a proper naturalized Sofia-citizen

• Went to bed early for once.

Conclusion

And that, dear reader, was The Bridge V.

If you didn’t like this article, it’s probably because you were not invited.

But don’t worry- big things are coming for The Bridge and our network will only grow stronger!

✌️

Founders, VCs, angels:
Our fundraising materials for raising $4.5M in San Francisco as Silicon Valley outsiders.

What it took?
85 ‘NO’s and a physical breakdown.

The steps?

Prep

1. Read books, start with

   1. Venture Deals (the Bible)

   2. Fundraising(short and sweet)

   3. Angel (very fun to read)

2. Meet founders

3. Prep materials

4. Get angels

5. Get VCs

The VC bits

1. Create an investor list

2. Get warm intros

3. Pitch deck is used to get a 1st meeting

4. NEVER show the pitch deck in the first meeting

5. 1st meeting is to get a 2nd meeting

6. Follow-up with the f*cking best materials

7. Receive a term sheet or get a ‘NO’ ASAP

Unfiltered, 1-hour lecture

Other materials:

1. PPT presentation from video

2. Team-GPT Data Room - Seed

3. How to raise $5M-$15M from Silicon Valley? Learnings from a conversation with a VC from Sequoia Capital

4. How Team-GPT raised $4.5M- A True story - 100% real story about our fundraising. Journalled as it was happening
   

Just turned 30.

Here are 20 things I realized in my 20s:

1. Consistency is the key to everything meaningful in life. I’m not saying never give up. I’m saying be consistently excellent at everything you do.

2. Aim as high as you can but also be patient to achieve your goals. But aim really high.

3. Luck is extremely important, but it’s not all ‘pure luck’. You are forging your own path and stumbling upon the luck on the way.

4. No one has time for anything. People MAKE time for what matters to them.

5. Explore as many different cultures as possible and try to integrate them in your own mindset. Integration is the key, not travel.

6. Experiment as much as possible while younger. Failing is great... if you do it as soon as possible (incl. as young as possible).

7. You will live in the same body, hopefully for the next 50-100y so start taking care of it. It’s not that hard – just be physically active every day and stretch often.

8. Start investing a % of your income in S&P500. Do this every month for the rest of your life.

9. Don’t plan about the next year or two; think in terms of the next decade or two. It’s hard when you are 20, gets easier around 30.

10. Meticulously remove toxic people from your day. Spending time with the wrong people is the root of all unhappiness.

11. Be a Yes-man as much as possible but learn to say ‘NO’ to protect your time.

12. Working in sprints allows you to dig much deeper into whatever you are doing. A normal sprint is 2 weeks and is followed by some chilled time. I stretch this concept to sprints that are MONTHS long and reach great depth of understanding. But I always chill well afterwards.

13. The ancient Greeks and Romans were right about everything! Try to strip down everything to its fundamentals and unveil its true essence.

14. Some decisions are irreversible, so think hard about them. Most decisions are reversible. When so- decide fast and move forward. Fast decision making compounds like crazy.

15. Chances are that for every single thing, there is someone, somwhere better than you at it. What sets you apart is the combination of your skills and experiences.

16. Your colleagues are not your family, instead you are like a sports team. You should feel comfortable passing the ball to them and trusting them 100%. It’s about trust, not entitlement.

17. Your family is your family. When all else goes away, that’s the only thing you have left.

18. Find someone smarter/ better/ more experienced than you to be your mentor. Mentorship is the fastest way to grow.

19. To truly change something about your behavior, it’s not about ‘changing your mindset to change your life’. Practically, it is about breaking existing habits and patterns.

20. You can always make more money, but you cannot make more time.
    
    🖖

‘How are your investors ok with you posting all this on LinkedIn’ - a friend asked me.

‘I guess that’s why they backed us in the first place’ - I replied.

Here goes our latest investor report.

January was about planning out 2025, doing some R&D, exploring the Saudi market, and shutting down v1 (effing finally!).

As of 1 Feb 2025:

• ARR: ~$415,800 (all time high)

• #Users: ~44,161*

• Money in the bank: $ 3,710,831.08

• Burn rate (approx.): $197,011 (spent in Jan)

• Runway: ~19 months (but starting to explore Series A to grow faster)

• # people: 25

* Following the migration, we found out there are ~5k users who’ve deleted their accounts or have not authorized us to move them to v2. We’ve decided to drop these ~5k users from our reporting. With them, the total # users, would have been 49,573.

Highlights

2025 Planning

We spent 1 week together in Sofia, planning 2025. Our goals: Revenue: $1M ARR by June 2025$3M ARR by end of 2025

1st Internal Hackathon

We split into 7 teams and had 2 days to built something cool. Crazy ideas were incentivized. The rule was to have something ready to be shipped as a tool in our experimental tools section. You can think of this hackathon as our R&D, which will be shipped bit by bit. Here are the results + our analysis of what was super smart about each tool and what wasn’t.

👉 1st place → YT Video Navigator. Spoiler alert, my team won after an audience vote. I hope it was because we built the best mini-product, not because I’m the CEO…

👉 All others ideas/tools in no particular order

💡 Email follow-up UI tool

💡 Excel Summarize and Compare

💡 AI notetaker

💡 Guided prompting mode

💡 Document filling assistant

💡 Content forger

💡 Code sandbox tool (similar to Claude artefacts)

💡 Generate Mermaid

💡 Speech-to-text tool

You will be able to use these tools in the ‘Experiment tools’ section in Team-GPT. Once we see traction in any of these, we will graduate them to the main bit of the platform.

In Jan 2025 we shipped

🧠 Project knowledge

• How to use Project knowledge → Guide

• Some people call this ‘Long term memory’. Other people call these ‘assistants’ / ‘custom GPTs’. Our ICP really resonates with the non-technical terms ‘Projects’ and ‘Project knowledge’

• Project knowledge is our take on reusable context. You can add project information (custom instructions), links, PDFs, CSVs, Excel files. Soon you’ll be able to connect different integrations (e.g. Google Drive, Sharepoint, Notion, Confluence, AirTable, etc.)

Project permissions

Completely rebuilt our permissions system to make it much faster and futureproof.

Improved model support

Gemini and Perplexity improved integration and support

Web browsing

One of the biggest objections we had was the lack of web browsing. Now that’s out, too.

Task management

Came out of a sales call, where we wanted to close a big deal. Interestingly, some companies consider chats = tasks. We are now gathering feedback on how this is perceived.

All release notes here

v1 is dead; long live v2

We finally ended the whole saga around Team-GPT v1. If we learned something from it, it’s: ‘don’t do complex data migrations, unless you really, really have to’. We consider this a mistake, especially so early in the development of our company. On the flip side, we managed to do it and people have all of their data migrated from v1 to v2. Hope this saved some valuable clients. We hope this will increase their loyalty and stickiness.

Why was v2 needed? Well Team-GPT v2 is enterprise ready. ISO and (soon) SOC-2 compliant and all the goodies. Plus, we’ve fixed all the architectural mistakes that we did in v1, which means we are now developing 5-10x faster.

Leaving some legacy behind:

• 47K sign-ups

• 1,473,208 lines of code

• 5195 commits

• 666 pull requests 😈

• 12 contributors

We went to Saudi to explore the market

The market looks extremely promising. We believe that given the Saudi Vision 2030, together with our strong connections there, we might have a good shot at becoming a market leader in the whole Middle East. We can get solid government support: $1.5M (RELOCATE) + $5M (Venture debt).

Read the memo here (link removed as this is confidential). Read the public version here:

We will only take next steps if they are revenue driven (e.g. we close revenue).

Deals overview

Asks

INTROS TO CLIENTS!!! Our product is now in a great shape, we’ve built a solid team and are ready to take care of the profits. We can close the deals, as long as they come our way. Please send us some deals!

Do you know someone who is figuring out AI adoption at their company? Just introduce us. We can help them!

Some collateral:

A PDF which you can check/send: Team-GPT Product and Company Presentation

Case studies (we will be improving these!)

👉 Marketing agencies (Die Crew)

👉 Consulting (SaaS Group)

👉 Software developers (Enlighten designs)

Priority companies to pursue: Marketing agencies, Consulting, EdTech.

Ideal size: 50-1000 FTEs.

That’s all folks!

Best,

Iliya

P.S. Original looks like this:

Leading to the Christmas holidays, OpenAI made a 2-week event: 12 Days of Shipmas.

Ship-mas is a pun coming from ‘shipping product features’.

‘Ship’ is a verb meaning, ‘send’ (your order has shipped). In the context of software development it means ‘put into production’.

Being successful in tech is all about shipping fast and OpenAI knows it.

The dust has kind of settled, so here’s the brief:

DAY 1 - o1 out of preview, ChatGPT Pro launches

On the first day, OpenAI launched ChatGPT Pro at $200/month.

This was not unexpected, in fact many in the industry experts expect that a license could cost $2,000/month in the future, given the more powerful models and AI agents.

The pricing reflects the substantial computing costs behind advanced AI models, particularly GPT-4, o1 and above.

DAY 2 - Open AI’s Reinforcement Fine-Tuning Research Program (alpha)

They opened up their fine-tuning capabilities to researchers.

While the market is rather small for this, it’s where true deep tech innovation happens, so it’s very strategically aligned.

DAY 3 - Sora and Storyboard

Sora and Storyboard are practically the next version of DALL-E (the image generation).

It’s created by the same team and allows you to create 10 second clips.

It’s really, really cool, yet super computationally expensive to be done well.

I still can’t get an account there- the demand is very high, while not enough servers are available.

DAY 4 - Canvas

Canvas was introduced as text editing solution. It looks somewhat like Notion but has pretty neat features for coding. It’s a take on Claude’s artefacts, while having a character of its own.

This validated what we at Team-GPT built 1 year ago (Pages).

Great minds think alike (or more probably - follow similar user data...).

DAY 5 - ChatGPT Siri Integration on iPhone/iPad (Apple Intelligence)

Siri finally makes sense.

Kind of.

Apple’s ecosystem is probably the most powerful distribution network for ChatGPT.

Now you can have ChatGPT on every iPhone in the world... except those in Europe of course, where regulation is still stiffling innovation! 🤦

DAY 6 - Video to advanced voice mode

They enhanced their voice capabilities with advanced WebRTC integration, paving the way for cool AI in real-time video communication.

When building my previous startup (3veta), we we creating a WebRTC project from scratch. This was some pretty complex programming, which tooks us 11 months to go to market.

Now we can have AI stuff inside the video. How fast these things evolve!

Expect advanced AI in Google Meet and Zoom very soon!!

(maybe I can finally stop attending all these meetings...)

DAY 7 - Projects for ChatGPT

Projects for ChatGPT launched, though several companies including ours had already shipped similar features. Great validation for the industry’s direction - I think the right one- good organization and logical data sources.

DAY 8 - ChatGPT Search

Head to head with Google.

Finally.

The whole world has been waiting for someone to contest Google its supremacy. This is ChatGPT’s Search functionality.

Problem is that it is built on Bing’s infrastructure (and we all know that we all prefer to ‘just Google it’, rather than ‘just Bing it’).

OpenAI is after the biggest players on the market and it’s not shy about it!

DAY 9 - o1 reasoning model into API

While it seems that OpenAI wants full supremacy over the users, in fact it wants full supremacy over the models.

They are trying to have the most capable and powerful models- and they are selling them through the API!

Last month our bill was $10K - paid to OpenAI, so they know how to monetize their models.

With o1 into the API, this is just showing how OpenAI depends on the developer community and AI startups to fill in the holes in its budget.

DAY 10 - ChatGPT calls and WhatsApp

OpenAI launched WhatsApp integration and voice calls, potentially reaching 2.5B users in regions with limited internet.

15-mins of talking are free per month.

I’m surprised OpenAI beat Meta to their own platform (WhatsApp).

DAY 11 - ChatGPT desktop apps

Desktop apps arrived, showing OpenAI’s multi-platform ambitions.

These were out before but my team is sceptical - it’s the closest to ‘spyware’ that exists. A software, listening to you mic, recording your screen, sending requests without you asking it.

Crazy!

The security concerns are out of this world...

But hey- who cares...

DAY 12 - o3 announced

The final day brought the announcement of o3.

o2 won’t exist because of potential copyright issues with the mobile carrier o2.

Apparently OpenAI’s o3 is still not available but it is bigger and better than anything we’ve seen.

Here’s GPT-4o vs o1 (the best model so far).

Now here’s o1 vs o3.

It’s too new to have better comparisons and you don’t need to understand these graphs at all to know that the newer models are much better.

The rate of improvement is incredible!

Conclusion

People often asks us: ‘How is Team-GPT doing when things are changing so fast’.

The reality is that with each improvement, the industry is showing that this is not just hype.

AI is the real deal and it will enter every part of our lives in the next 5-15 years!

I’m beyond excited to be building in the Era of AI.

Date: 16 Dec 2023

This is what I said to a prominent venture capitalist from Sequoia Capital- one of the most successful funds in the world.

We were at his house sitting on the table, where I gathered the most crucial information for my fundraising strategy. I knew it was unlikely for this person to fund Team-GPT. So I went on to ask him a bunch of questions.

I was taking notes the whole meeting.

This article goes into all the questions I asked and the answers this VC gave me.

I’ve heavily based the fundraising process of Team-GPT (raised $4.5M) on this information.

Here are the questions + key findings.

What key metrics or milestones should I focus on?

• Show constant growth

• Paint a big picture and tell a compelling story

• Focus on how this becomes $500M ARR in 10 years

• The more metrics you have, the better

How to target investors?

• Talk to founders and get warm introductions to VCs

• Start having informal meetings with investors way before you fundraise

• Develop relationships now, as most investors are in for the next 10 years

• Be prepared for “We are not investing yet but...” conversations

What are realistic amounts of capital to raise in a first round?

• For first-time venture-backed entrepreneurs, $15M is likely too much

• Start with Seed and target $0-$3M.

• It’s supply and demand. If more people want your deal, you will likely get a better price

What kind of due diligence should I expect from investors?

• You should know all the common ones by heart (ARR, #users, MAU, etc.)

• Do cohort analysis

• 30d, 60d, 90d retention of different cohorts

Who leads the round - does it matter and why?

• It matters significantly

• Every lead investor in Seed optimizes around ownership

• Whatever you raise will likely cost 20% of the company

• Everybody hates co-leading (10% for one investor, 10% for another doesn’t make sense for anyone)

How to choose lead investor?

• Main decision: “Do we like you, do we like the idea”

• For Seed round, look for: Outlier founder (most important) with Unique and compelling insight in an Area where you can build a large business

• Ask yourself: What investors do you want to have on the cap table?

Timeline for raising?

• Be conservative, 1-6 months, brace for 3 months

• Most time-consuming part is getting to the first term sheet. You just need one.

Thoughts on Business model?

• Don’t set your price low

• Aim for around $50/user/month

• The moment you set your price, you set your target market

What are the key considerations for the Board of Directors?

• It’s Seed round - YOU (the CEO) are the board.

• OpenAI’s recent issue is just one of many examples of board-related problems but these will come later

Where should the company be incorporated?

• Incorporate in Delaware. Full stop.

What are some examples of Sequoia’s recent investments?

• Sequoia just invested $15M in a company that doesn’t have revenue

• Why? The founder had several successful exists that made money to their investors

Examples of people who can raise $15M on an idea:

• Ofir Ehrlich (LinkedIn profile: https://www.linkedin.com/in/ofirehrlich/)

• Assaf Rappaport (LinkedIn profile: https://www.linkedin.com/in/assafrappaport/)

Additional note:

• Wiz (company founded by Assaf Rappaport) raised significant funding from the get go

• Sequoia is willing to invest large amounts ($15M) in pre-revenue companies

• Founder track record is highly valued (multiple successful exits)

• Successful founders can raise substantial amounts even without current revenue

How does Sequoia approach investments?

• Sequoia makes about 20 investments, totaling $700M

• 10% of these investments drive the returns

• They invest with very high conviction (translation: they need to be super-duper sure you are a winner!)

How should I frame my pitch? The VC advice on Team-GPT:

• Don’t lead with “ChatGPT is difficult to adopt” as it won’t resonate with investors. ChatGPT is perceived as a good product

• Focus on the transition from single player to multiplayer

• Emphasize the shift from personal to company-wide story

What narrative should I build? The VC suggestion:

• Enterprise communications have only been people-to-people

• Going forward, it will be people-to-people and people-to-machine

• There needs to be a new enterprise chat system that mixes both

What should I emphasize about myself?

• Focus on why I am the person who will build this company -> you are the CEO and have to show you are ‘calling the shots’

• Sequoia would invest in companies based on “Act 1” (the founder’s background and vision)

Key analogies for Team-GPT:

• Slack is great for human-to-human communication

• There needs to be a new enterprise chat system that incorporates both people-to-people and people-to-machine interactions

How should we structure the initial equity distribution?

• In the beginning, founders are getting everything

What about early investors?

• It’s better for me to have preferred stock than common stock -> later it turned out that founders always get common stock

How much should we allocate for the stock option pool?

• Stock option pool should be 10-15-20% of the company

At the end I had written Next Steps

To do:

1. Every week work on the deck a bit more

2. Focus on the storytelling, not the numbers

3. Metrics -> engagement and retention

4. Cohort analysis - 30d, 60d, 90d retention of different cohorts

5. Talk to founders that have raised before

Talking to other founders

Indeed talking to founders was crucial.

I set a goal to meet 25 founders.

I met about 15.

These founders told me everything I needed to know about the fundraising process.

The Sequoia VC told me that books will be a waste of time.

I disagree then and now.

Read these 3 books before starting to fundraise:

• Venture Deals

• Fundraising by Ryan Breslow (short and sweet)

• Angel by Jason Calacanis (very fun to read)

They’ll help you think like a VC!

Thanks for reading and good luck fundraising!

I’m a serial entrepreneur who has founded 10+ initiatives and 3 startups. These have reached ~3M paying users throughout the years.

When people say – it’s hard to get from 0 to 1 - I disagree. You just have to validate the idea well.

I’ve done it several times, but no other project has had a stronger start than the current one: Team-GPT.

• 9.5 months

• ~$200K in revenue

• 25K users

• 3K active businesses

• Used by: Maersk, Johns Hopkins University, Yale University, Shift4, Charles Schwab, and more.

I wrote a bit about this in the From 0 to 10K users and I suggest you start there.

Here’s how we did it.

The experiment

GPT-4 came out, we tried it, and we were crazy about it. For a couple of days, we tried to incorporate it in our work but talking to ChatGPT was a very personal process: always 1-1 with the AI.

We really hated the lack of collaboration as it was hindering out ability to adopt the technology as a team.

With a strong product team already in place, it took is a couple of days to come up with an internal tool, which allowed us to collaborate with the AI.

The idea was simple: collaborate in chats, organize in folders, save prompts for future use.

I had just read the book ‘Pretotype it’ by Alberto Savoia (first engineering hire of Google). It contained a strong idea: the XYZ hypothesis, which states you should test product ideas with: ‘At least X% of Y will do Z’.

As a data scientist with 1.3M AI students, I loved this: a hypothesis testing approach for validating product ideas! Let’s put it to work!

Our initial hypothesis?

Once you have a hypothesis, it must be tested with real clients:

1. Create a website (landing page)

2. Add a payment option

3. Promote it to someone…

Who do you promote it to?

We had several targets in mind:

• Marketing teams and agencies (who create a lot of content)

• IT teams (who write code)

• AI startups that are very open to AI tools

At a price point of $29, paid one time – it sounded like a no brainer.

And it was!

We hit $300 in sales in the first two weeks…

Biggest surprise?

We didn’t even promote it. People found us organically on Google. We had a simple WordPress website, optimized for the words ‘collaborate in ChatGPT’, ‘ChatGPT for teams’ and several others.

And yes, we got lucky with the name (team-gpt.com cost $10 back then), and I understand this will not be the case for many others – but this is our story, and I can’t change it. I’m just grateful it happened.

They say ‘build it and they’ll come’ is the wrong approach… because they never come. Usually true.

In this case, they came before we built it.

There was no product, yet we were making sales. There is no stronger indicator that you should build it.

Soon, we raised the prices to $39/lifetime, $49/lifetime, $10/team of 10/m.

Right now, we charge $2/user/m, charged in bulk for 10 people, so $20/m for a team of 10, scaling up to $1000/m for a team of 500. The pricing is simple, yet effective.

While there was the potential to charge even higher prices, we decided to keep growing the user base at a lower price point.

Product development

After the first clients came, we had 2 choices:

1. Refund them and start thinking about the next steps

2. Propose that we refund them and build it immediately

We asked… and the people wanted the product, not the refund.

So, we built it on top of the internal tool.

We used a tech stack of:

• An open-source project that we started building on

• Clerk for client authentication (this literally brings the ‘team plan’ out of the box)

• MongoDB Atlas (a scalable cloud database that allowed us to take any number of users)

• Lemon Squeezy for payment processing

The first version was out in a couple of days, built entirely by Ilko and Yavor (the other two co-founders of Team-GPT), and it was a system with many bugs, broken interface, practically no mobile version, and some of the worst tech debt in the universe.

But a good old startup saying goes like: ‘If you don’t hate the first version of your product, you launched too late’.

The product was working and most importantly, it was solving a big pain: collaboration in ChatGPT.

9 months later we’ve got a great system with hundreds of features, based on customer feedback.

We’ve also built a separate B2B app, that caters to Enterprise clients for whom the initial tech stack was suboptimal.

And remember - you can always change the tech stack if you need to. Just don’t overkill from the start.

Wouldn’t ChatGPT build it? They did!

Truth is – every business is a limited time opportunity. Then it iterates and reinvents itself.

I had just read the story of Ruslan Leteyski, who had made $10M+ from an app that enhances the checkout experience in Shopify. They reached $600K ARR before Shopify shut his operation down and that’s a pretty good outcome.

This showed us that no matter if OpenAI does it in 5 months or 2 years, there is no point in us missing out on this opportunity.

And we knew ChatGPT would not copy us – they have their own plans, nor would they shut us down – we were bringing fresh users, paying for their services. In fact, we were capturing a segment which OpenAI was not serving, making them money, and popularizing the technology. We were not a threat – but a partner (users are paying OpenAI directly for the model usage). Every other business built on top of GPT models was in the same boat as us and that’s why the OpenAI APIs exist.

It took OpenAI 9 months to release a team version and it is still pretty basic. Here’s how ChatGPT for teams compares to Team-GPT.

This gave us 9 months to think about what we want to pivot (if we needed to).

Was a pivot needed?

We were well aware that we can’t be just a collaboration wrapper on top of ChatGPT.

3 days after we released the first version of Team-GPT, I was already on my way to San Francisco, going around meeting people. I met with people from OpenAI, Google DeepMind, AI startups, literally anyone who could give me information. The main question?

‘Where is this going, what does the future look like for ChatGPT’

After realizing that ChatGPT’s success was a shock even for OpenAI, I knew they weren’t going to rush to compete in this SaaS realm. Time was on our side. Google was catching up in terms of models. Amazon invested $4B in Anthropic – to compete on the foundational models, too. The world was on fire and the big players did not really care about the small gains of SaaS software.

They all cared (and still do) about AGI.

In the last 9 months we’ve been conducting many customer interviews and focused on building the best experience for our users. We listened to their feedback and we incorporated it – and they love us for it!

You read that right – adoption was the key.

Adoption is the key

For a product to be used – it needs to be understood and adopted.

What could we do better than OpenAI?

Education.

Based on our team’s previous experience, we knew we had to bring something more to the table – and we did.

We focused our whole product around adoption:

1. The online course

I’ve created 30+ courses, with some of them having 600,000+ students, 130,000+ reviews (see the The Data Science Course).

So, we rushed to build an online course. I disconnected completely for 2 weeks and did what I do best – an online course.

ChatGPT for Work: The Interactive Course was built in record time (thanks to using ChatGPT of course).

And we released it for free.

A free online course, built by a top instructor - all inside the software.

Every new user of Team-GPT would go on the platform and be urged to take the course.

And when they take the course?

They’ve learned how to communicate with the AI properly.

Not only this.

They’ve learned how to use Team-GPT.

And since the product is rather good – they simply stick with it.

This became a major part of our marketing efforts.

Users come for the course, stay for the software.

Or… they come for the product. They stay because we taught them to talk to the AI.

Many of our users prefer Team-GPT over ChatGPT because we helped them see the value.

And that’s powerful.

2. Prompt library

Nurturing the usage was key from Day 1. How do we help users prompt better?

By preselecting useful prompts for them and providing them for free as a part of the product.

The prompt library kickstarts the users’ prompting journey and supercharges their prompts.

The feedback we got was that people also use our prompts to learn to prompt better (more education).

Once they are proficient, users can also create their own prompts and store in their ‘Workspace library’.

3. Tips and tricks

Each chat boasts tips and tricks in the middle (see above).

Every time you start a chat, a tip is shown which further helps your learning.

The more you use the product, the more tips you see (it’s like a computer game).

4. Adoption reports

So, we are very good at adoption (nurturing the usage of the software).

How do we know? We measure it.

And since we were already measuring the adoption, we realized there is nothing better than showing these reports to our users, too!

This greatly helps all admins and managers nurture the adoption of the software throughout their organizations.

Much of our adoption success is due to our ability to make users use Team-GPT.

And this is where we are so aligned with our clients (the businesses).

Everyone knows ChatGPT is a real deal.

But Team-GPT cares about adoption.

And adoption is exactly what companies want at the end of the day. Not only to pay for the software but to make sure all employees are AI-enabled.

User acquisition

While user acquisition is crucial for a 0-25K users, I’ve covered this well in another one: From 0 to 10K users | Team-GPT Getting Ready to Scale, so I’ll spare you a lot of it here.

In fact, most of it was carefully following the plan outlined in this 16-min onboarding video for our sales and marketing hires: Sales and Marketing Strategy: How to Achieve our Revenue Goals📈

We are experienced marketers, and we know there is 1 thing that’s most important: MOMENTUM.

All marketing campaigns need to follow one after the other.

And when they are planned well, they create a chain reaction.

My favorite channel? The one that brought the first users: SEO and content marketing.

We saw the opportunity and grabbed it.

We’ve invested more than almost any other early-stage startup in SEO and this was a bet, which is paying off tremendously.

Today we are getting 1,500 clicks and 25,000 impressions on Google every day.

We are ranking for thousands of words and are on first position for many of popular ones:

In fact, ChatGPT releasing their team plan has not hurt us.

On the contrary, they’ve generated demand, bringing even more traffic to us.

Social

Once you have so much content and you are good at it, it is a shame not to promote it all around the place.

Our main success is on YouTube, Instagram and LinkedIn (this article included).

We’ve published hundreds of short videos.

The rule us: reuse, reduce, recycle - or as we call it in marketing: repurpose content.

We have repurposed so much of our content and are posting it all around the place.

Here’s what our YouTube metrics look like.

Instagram and TikTok are hit and miss but we did have a fair share of viral videos, reaching 6,752 followers.

We don’t spend much time on this, we just publish content we already have.

Finally, LinkedIn.

This one takes the most time for me, but it is crucial for our long-term success. LinkedIn is the professional network where most of our users check if we are real people – and that’s really the case when you sell to businesses.

I’ve been regularly posting for about 1 year and here’s how many people I’ve reached.

Now LinkedIn has put more eyeballs on our stuff than YouTube and Instagram!

I’ve never had a viral post (my best posts get ~15-20K impressions, while the average is ~2K).

I’m not in a rush to grow faster - I keep the content relevant to our audience and align it with the business goals.

Retargeting

Retargeting is extremely underrated.

I don’t know why people think ads are expensive.

Ads out of the blue are.

But once someone lands on your website, it is rather cheap to show them an ad (retarget/remarket them).

We have spent a bit below 920 EUR to bring back 41K users and show our ads 5.5M times.

We are spending around 5-10 EUR/day for 10K impressions, which are RELEVANT to the people we are targeting.

We are aiming to show a certain ad < 5 times per person, and we’ve got 30 variations that we shuffle every 10 days.

The main point: make people remember you, without being annoying.

25K users: Digging in the data

One of the biggest questions is not how many users you’ve got.

It is how active they are.

First – how do you measure it? We are using June for this – took us a couple of hours to set up and the rest has been a breeze. It’s a bit simplistic but does the job.

Second – what do you measure?

When it comes to GPT usage, the relevant metric is WEEKLY BUSINESSES. We (and OpenAI) consider someone an active ChatGPT user if they have used it once per week.

Our weekly active base is growing fast, showing that they actually use the software.

Right now, we’ve got around 3K weekly active businesses using Team-GPT.

The top use cases, based on self-reported data:

• Marketing

• Education

• IT

These are the use cases where people immediate see the value from using GPT models. The early adopters of the GPT technology.

The Enterprise opportunity

Do you remember when we spoke about what we would do if OpenAI introduce ChatGPT for teams?

We had another hypothesis.

Enterprise clients would not want to share their data with OpenAI.

This is how we built Team-GPT Enterprise – the ability to deploy Team-GPT on premises or on private cloud and hook it up with any model (incl. Microsoft Azure, Google Bard, LLaMa, Mixtral, and more).

As teams from bigger companies were starting to use us, we were getting the following question:

‘Do you offer a private deployment of your software?’

And yes – we started offering it.

Enterprise deals are not easy, though.

We have a pipeline of 20+ companies that are seriously interested in the product above, with 10+ inbound (SEO-driven) inquiries each week.

But it takes time to convert and Enterprise client – 6-12 months on average.

The good thing?

While they are considering the Enterprise option, we hook them up with the ‘$2/user/m’ product. Today we’ve got 20+ companies with 60+ active users (4 with 250+ users), negotiating a private instance to switch to.

We know that the big opportunity lies in Enterprise, and we are gearing up to scale our sales efforts. The negotiations are never easy but once they convert, we will be into the serious business of Enterprise software.

The future: Time to seize the opportunity we’ve been given

I’m a big optimist about the future of Team-GPT… as every startup CEO should be.

All signs are positive, and we’ve done most of the things right (many mistakes were made but much less than with my previous startups).

We’ve got a strong team.

It’s getting easier to acquire new users.

It’s also easier to attract top talent.

We know where the opportunity lies – and clients are finding us on Google.

Our next goal? Find the right funding partner for our big plans. We’ve got a the core team in place and are ready to receive a big financial boost.

10 years from now, most work will be done by AI agents, while humans set goals, monitor and approve their actions.

Team-GPT will be the place for this human-AI collaboration.

🖖

Thanks to everyone involved!

6 Nov marked OpenAI’s first conference: OpenAI DevDay. It was the biggest event in AI.

I was there and lived to tell the story!

Sounds dramatic but many startups died that day, and we were not one of them.

Here’s my honest review of what was announced and where OpenAI is trying to position itself.

Here’s what we’ll cover

• What was announced at OpenAI DevDay

• Several mini-announcements with major impact

• OpenAI + Microsoft

• GPTs

• Assistants API

• Evening reception at the de Young Museum.

• After party with Grimes

• Summary

Who was at OpenAI DevDay

The name itself: ‘DevDay’ was showing it was aimed at developers.

I could count ~1000 people at the event and most were not developers at all.

Based on the sample of the people I met, I’d guess the distribution was the following:

1. 1/3 OpenAI team or companies closely affiliated with OpenAI (e.g. ChatGPT Enterprise distribution partners, BCG consultants, freelancers doing stuff for OpenAI)

2. 1/3 students or recent graduates. I’m unsure why there were so many of them but I can imagine this was a recruitment technique

3. 1/3 miscellaneous. I was lucky to be among those invited to OpenAI DevDay and I think I fall in the misc group. Among the rest of the people (like me) I could not see a common thread – they were anything from educators, through researchers, through entrepreneurs – a mixed bag. The main thing we had in common was that we were not in the first 2 groups.

To gossip a bit - among the people I saw that day were: Sam Altman (CEO of OpenAI), Satya Nadela (CEO of Microsoft), Greg Brockman (President, Chairman, & Co-founder of OpenAI), Grimes (world-famous for being mother of Elon Musk’s children but highly appreciated in Silicon Valley for all her other talents) and many others famous in the AI community.

The little fanboy in me was very excited!

The venue was probably the most convenient conference venue I’ve been to – it was on several levels, while one of them (where the food was) was completely outside (OpenAIR they called it). Kudos!

Schedule

08:30 AM – Doors open & breakfast is available
10:00 AM – Keynote begins
11:00 AM – Lunch & demo lounge opens
11:30 AM – 12:15 PM New Products Deepdive
11:30 AM – Breakout sessions begin
12:30 PM – 1:15 PM Maximizing LLM Performance
01:30 PM – 2:15 PM The Business of AI
02:30 PM – 3:15 PM AI Frontiers 03:30 PM – Content ends
05:00 PM – Evening reception at the de Young Museum

We went in around 8.30AM, had breakfast and networked until 10AM.

At 10AM Sam Altman held the keynote.

Right after the keynote… most of the people disappeared. There were 2 stages with events going on all the time but those were mostly empty.

It was such a Bay Area experience – people came for the keynote and went to work, I guess? Who does full day events on a Monday anyways…

There were also the ‘Breakout sessions’, where OpenAI was getting ideas from the community on what to do next. Many of the important meetings happened in these breakout sessions behind closed doors.

This involved mainly the first 1/3 of the people at the event.

What was announced at OpenAI DevDay

By now a lot of this info has reached you but in this analysis, I’ll not only present the information but also what it means for OpenAI, Team-GPT (and companies like us), and the whole ecosystem.

These impressions are based on 150+ conversations I’ve had between 3 Nov and 9 Nov in the Bay area.

The success of ChatGPT

The day started with a celebration of ChatGPT numbers: 2M developers, 92% of Fortune 500 companies, 100M weekly active users.

1. 2M Developers. What does developers even mean? It means you are using an API key to do something with OpenAI. Could be to build an app, integrate LLMs in your existing product or simply chat through the API. Chances are you (dear reader) are also a developer according to this measure.

2. 92% of Fortune 500. This number sounded like a playbook from Sam Altman’s YC past - a rather ambiguous number, with not much context which sounds very impressive.

3. 100M weekly active users was the highlight for me.First, because I do believe ‘weekly active users’ is the right measure for ChatGPT usage (at Team-GPT we use the same). The reason is that we are not quite at the point where people use ChatGPT every day. Judging by our own users, 2-3 times per week is the norm. Some tasks just can’t be ChatGPT-ed yet or we haven’t found all the use cases. But soon GPT usage will become daily – similar to how people adopted Google. First for specific use cases, now Google is literally ‘the front page of the Internet’.

On the topic of users, I want to elaborate a bit more.

100M sounds like a lot but it isn’t. Nothing was said about GPT-4 usage vs GPT-3 usage during the conference but crosschecking with other sources, only 2M of those are using GPT-4 (which is the real deal).

GPT-4 = paid users who are getting the actual value out of the product.

2M weekly active GPT-4 users out of 2B people who will need to use ChatGPT for their work in the next couple of years.

2M / 2B means we’ve got < 0.1% GPT-4 adoption.

99.9% of the population is not using GPT-4.

Even if we count the 100M that Sam Altman quoted, humanity is at <5% adoption (if we assume that these 100M have really ‘adopted’ ChatGPT).

Yet not a single word was said to address this growing gap between those who have adopted and those who haven’t.

GPT-4 usage gap means we’ve got AI-enabled humans, competing with simple humans.

We talk about the ‘gender pay gap’, ‘race pay gap’, etc. Those would be NOTHING compared to the AI pay gap. People using AI can and will make 3x, 5x, 100x more than the others.

And at the OpenAI conference and in San Fransisco as a whole, this whole ‘AI-adoption gap’ was rather funny as a concept – another way in which people in tech are ‘superior’ to everyone else.

If you are a part of the ‘everyone else’ group, don’t be antagonized by this statement. Instead, start using GPT-4 immediately and catch up. I’ve created a FREE course: ’ChatGPT for Work: The Interactive Course’ to help you with this.

Actually, the only time we addressed this was in the Uber going from one venue to another. Over there, a kind Mexican lady suggested we share the Uber – something I had never seen in America before! Either way, she understood it because she was not working in the tech industry, she was working with tech industry.

Several mini-announcements with major impact

Most other articles on the topic start here. I’m sure many of them will be more detailed but I’ll try to add my perspective on the impact of each of these.

1. GPT-4 Turbo and 128K Context Length

Woah! This was huge!

As Sam Altman put it – you can now feed 300 pages at once as context.

The main GPT-4 competitor is Anthropic (with its model Claude). Claude had one big advantage over ChatGPT, it could take up to 100K tokens as context (~75K words), while GPT-4 was limited to 8K tokens (~6K words).

With this move, OpenAI literally annihilated Claude’s only advantage.

By the way, Amazon recently invested $4B in Anthropic.

This announcement must have hurt both Anthropic and Amazon, so I’m curious to see how they respond.

2. More control through the API

Function calling is now more advanced.

You, dear reader, are probably not that technical (as my audience normally isn’t), so I’ll just scratch the surface of this one.

OpenAI is allowing companies like ours to now create plugins, agents, seed (for reproducibility) and more.

This was one of the biggest hurdles for developers (like us) and these updates will allow us to build better products.

Thanks, OpenAI - with the caveat: all of this must happen within OpenAI’s ecosystem.

Not that sexy but we at Team-GPT are happy with this development.

3. Knowledge cut-off is now April 2023, not September 2021

The knowledge cut off is pretty annoying, right? But as Sam Altman put it: it is even more annoying to OpenAI.

When ChatGPT came out in Oct 2022, they were 1 year behind (data until Sep 2021).

During DevDay in Nov 2023, they updated the cut-off to April 2023, so they are now 6 months behind.

OpenAI is narrowing the gap, but LLMs (GPT models) do NOT have contemporary knowledge by design. This is also why many people don’t want to use them or don’t trust them. This is also the main advantage of Google Bard.

OpenAI has a fundamental problem with this and I’m curious to see how they go about it.

Sam couldn’t fool us on this one!

4. New modality

GPT-4 Turbo with vision, DALL-E 3, new text to speech all available through the API. This is generally great news for developers. Now we can do much more with the API and are happy to experiment!

DALL-E 3 is now the best image generation API out there and on the same day, Coca-Cola launched a campaign produced with the new DALL-E 3.

This was a real flex!

But the really big one for me was text to speech.

Check the Text to Speech of OpenAI over here.

It’s quite something – a text to speech which I would call almost perfect.

Soon available through the API through Whisper V3.

5. GPT-4 fine tuning + Custom models

This is where it becomes interesting.

Sam mentioned that GPT-4 fine tuning will be available – and they really had a lot of pressure to do so. Based on my experience, I know very few people who have had success with fine-tuning. Moreover, fine-tuning GPT-3 is rather pointless, since even the best GPT-3 fine-tuned model is still worse than the generic GPT-4 (with very, very edge case exceptions). Yet to see good implementations of this.

Custom models, however, are something else.

Custom models are apparently the possibility to hire OpenAI to train a custom model for your company. This sounds very exciting, until you realize ChatGPT still doesn’t offer a team plan, takes weeks, if ever to even schedule a ChatGPT Enterprise demo. But now they will be doing something which is very human-resource intensive (essentially consulting). This is when I laughed a bit because it could be translated as: OpenAI wants all GPT-related business and doesn’t care that they can’t handle it.

Custom models sound like the last thing they want to do if they want to reach AGI. Unless, of course, they train on the data of the clients for whom they are building the custom models. This is the only way in which I see them aligned with working on this.

This reminded me of another meeting.

I remember meeting with the Head of Growth of Hugging Face and asking them how we can work together. We briefly discussed the possibility of consulting and Custom models and he told me: we are doing this – we don’t need anyone to help us. FYI, he replied like this to almost everyone else who asked for partnerships: ‘we are doing this’.

In a similar vein, OpenAI tried to say the same thing during OpenAI DevDay: ‘We’ve got this’.

Although they don’t.

1 year later, they still can’t handle ChatGPT Plus sign-ups.

But they will be offering literally every service ever.

Interestingly, there were many ‘developers’, aka attendees/invitees to the conference who were very much in the business of Custom models. OpenAI is eating them alive, while not being able to provide the desired service for even the basic product to its users.

That’s what infuriated everyone, and we will revisit this again.

6. Higher rate limits + Copyright Shield

Ah, hallelujah!

This is literally what we want, OpenAI. All the heavy users are complaining about rate limits - everywhere, all around the place. This is the first thing I hear when I talk to clients!

They’ve doubled the rate limits to: 2x tokens (for existing customers) and created a procedure to increase your current rate limits (available here).

Here’s the quick overview.

Then again, 2x is nothing. If I’m hitting the limits every minute, then by doubling the tokens, I’ll be probably hitting them even with this 2x increase. This is not allowing devs to scale operations well.

We need 10x to 100x increase in the rate limits and that’s where Microsoft will likely jump in.

Again, all of OpenAI’s resources seem to be aimed at increasing the scope of their activities, rather than the scale.

Finally, the Copyright Shield… As Sam put it: ‘We will step in and defend our customers and pay the costs incurred, if you face legal claims or copyright infringement’.

This is just about everything he said, and we are yet to understand what this means.

Sounds like: OpenAI’s lawyers are my lawyers now – free of charge…

Interested to see how this scales - it’s either no one will need the Copyright Shield, or everyone will need it and the system will collapse.

GPT-4 Turbo Pricing

On to the good news – a major price decrease!

The newest model is not only the best one ever but also cheaper than the existing one.

When GPT-4 came out, it was 20x more expensive than GPT-3.

As GPT-4 Turbo is coming out, it is ~2.75x less expensive than GPT-4.

Everyone wanted cheaper models and OpenAI is providing.

But this sounds strange, doesn’t it?

OpenAI is currently leading a natural monopoly. Everyone knows GPT-4 (Turbo) is the best model. I’ve been preaching to everyone everywhere – do not use anything less than GPT-4, it’s not worth your time.

So why are they slashing the price?

I’ve investigated this at length in Sillicon Valley. I asked many people about their point of view and 2 major opinions formed:

1. LLM costs will be going down in the future and this is a part of the trend.

2. Defensibility - they are trying to lure more developers ASAP and maybe raise prices later

Both of these make sense and are probably true at the same time.

With the higher scale of operations + the partnership with Microsoft, OpenAI is probably able to slash costs 10x to 100x more, while allowing us to enjoy only 2.75x of it.

This is great news, and we are grateful for the gesture – they could have just continued extorting developers, being the natural monopolist they are.

But I want to touch on the second point – defensibility.

OpenAI has not only increased the context limit to 128K (above that of Anthropic), they are now also slashing the prices. This is making it even harder for others to compete. OpenAI is trying to destroy the competition on every level they can think of.

Given the lower prices, we are all thinking about integrating even further with OpenAI. But nothing is stopping them from changing their mind and increasing the prices 5x in the future. We have seen this in almost every industry ever, most notably cloud, where it is very cheap to get started but then hits very hard as you scale up.

Or simply doing something else…

For instance, pushing 128K tokens through the API at the ‘new lower prices’ amounts to about $1.28/request. And this is the beauty and brutality of the combination of the new features – they allow you to pass 16 times more tokens but charge per token. They give you more, so they can then take more.

One of the first pressure tests which came out on Nov 9 was by Greg Kamradt who found out the 128K tokens model is not that great but cost him $200 to prove it. Here’s the summary:

• No Guarantees - Your facts are not guaranteed to be retrieved. Don’t bake the assumption they will into your applications

• Less context = more accuracy - This is well know, but when possible reduce the amount of context you send to GPT-4 to increase its ability to recall

• Position matters - Also well know, but facts placed at the very beginning and 2nd half of the document seem to be recalled better. Facts placed between 7%-50% document depth are not recalled well.

Other than that, we can still use the old models in the same way we did.

All models are getting about the same price reduction as the one mentioned above. Same applies for fine-tuning GPT-3.5 Turbo.

And that’s only 15 mins into the conference.

A lot more happened later on, and we will go through it all.

I must say, the density of new information at this keynote was out of this world!

OpenAI + Microsoft

About a year ago, Microsoft invested $10B in OpenAI for a 49% stake.

Several months later, we saw Microsoft unroll a bunch of AI-enabled features, most interestingly – the co-pilot, an AI that is supposed to help you with everything (a story for another time).

OpenAI needed a partner – one with a lot of infrastructure. And while Microsoft had a lot of it, it announced it is setting up another $3.2B worth of servers in Australia alone.

This is important because they bought half of OpenAI for $10B. And now they’ll probably invest way more than that to make ‘ChatGPT happen’ because computing is expensive.

That said, Sam and Satya looked like best friends forever (a bit awkward but that’s our usual Sam)!

I must say – that was cool! Having Sam and Satya in one place was really powerful.

Satya was at the keynote for exactly 4 minutes, said a bunch of stuff that sounded ChatGPT generated, Sam dubbed it the ‘most exciting partnership in tech’ and Satya left.

A bit underwhelming in retrospect but REALLY COOL in the heat of the moment.

It was a strong PR stunt and OpenAI are really amazing at PR and marketing.

As a marketer - hats off for this move!

ChatGPT just became ‘dumber’ or why the model picker is gone

Sam began this with ‘we heard your feedback’ about how annoying it is to choose the model every time.

This means that now, whether you are browsing the internet, creating images or using plugins – it is all automatic.

Was it really that annoying? Who was it annoying for?

With the current updates, ChatGPT is now browsing the Internet when I don’t want it to. Creates images when I don’t want it to. Invokes plugins out of nowhere.

Overall, I think it became dumber. But the problem is not that it is dumber.

The problem is: my time from prompt to outcome is MUCH SLOWER.

People use ChatGPT to save time, not to waste time.

And that’s what they’ve achieved with this new feature.

Was this the customer feedback? Definitely not mine.

I remember telling 3 different OpenAI employees how their automatic choice of model is pretty bad. I’ve been providing this feedback since May 23, when I had a long discussion about Plugins.

Now this is my feedback to OpenAI: ChatGPT became much dumber with this product decision. I’m sure it will get better but right now – it is pretty bad.

I have certain information that they’ve put a lot of resources into getting this right and are obsessed with doing so. Yet I still think it is pointless.

I’ll let it slide for now because of an awesome new feature, which I think they got right.

GPTs

Another genius marketing move. Bravi!

What do customers want?

Every customer says the same thing: ‘I want to train my own ChatGPT’.

They took this customer request and introduced GPTs. As they described them: ‘GPTs are a new way for anyone to create a tailored version of ChatGPT to be more helpful… ‘

Here’s an example with an actual GPT that OpenAI launched: The Negotiator.

While I cannot know how exactly they prepared ‘The Negotiator’, my best guess is that it has some custom instructions to be very good at negotiation. Maybe they provided a couple of PDFs with negotiation tactics (but I doubt it – I think core GPT-4 is good enough).

Important: This is not a new model. It is the exact same model but with some differences to its capabilities and instructions. No fine-tuning was made. No new model was created.

What are GPTs exactly?

GPTs are something like ‘chat templates’.

These chat templates have several configurations which are set at the time of creation. Here are the possible configurations:

1. Custom instructions – this is what makes GPTs so good. Normal chats in ChatGPT all have the same instruction ‘You are ChatGPT a model trained by OpenAI. Respond using markdown’, or something close to this. These instructions are sent with EVERY message (we call this ‘system prompt’). You could change this inside the ChatGPT app using the custom instructions feature (see image below). In fact, using the API you could use custom instructions since 2020. Problem? No one really cared to use custom instructions, although this is probably the most important feature of ChatGPT. Through the new feature: GPTs, custom instructions are embedded in the GPT itself. For ‘The Negotiator’ from above, the custom instruction is something like ‘Act like a professional negotiator. You’ve been trained for many years to negotiate any deal from the smallest business to Enterprise deals with the toughest negotiators in the world’ or something of the sort. Note that millions of different wordings would work ‘good enough’ but you get the gist of it.

2. Knowledge (files, aka RAG) – Another big feature coming to GPTs. RAG (retrieval-augmented generation) is another important feature. It allows you to add files (PDFs, spreadsheets, and other file types) and use the information from them during your chats. While this feature was available before, GPTs allow you to upload files once and then reuse them in many different chats (e.g. The Negotiator can be used as a chat starter for many chats). Using RAG is also one of the best ways to deal with hallucinations. Therefore, if you configure your GPT correctly you might get rid of many of the hallucination issues, too.

This ‘knowledge’ that we speak of, aka file upload, aka RAG is optional. You don’t have to use any files if you don’t want or need to.

Same applies for the ‘capabilities’: Web browsing, DALL-E Image Generation, and the Code Interpreter

3. Web browsing – We all know that ChatGPT is not great at web browsing (because it uses Bing underneath. Hint: Use Google Bard if you want to search the web). Do we always need it? Not really. ‘The Negotiator’ doesn’t really need to browse the web. Therefore, OpenAI allows us to simply disable the Web browsing for a particular GPT (chat template) and thus save time and resources (both for the user of the GPT and OpenAI itself).

4. DALL·E Image Generation - Your GPTs can also generate images… if you want them to! If not, simply uncheck the box of this capability. I never use it. Frankly, I still prefer Midjourney, but DALL-E 3 is getting there.

5. Code Interpreter – This is what they call ‘Advanced data analysis’ inside ChatGPT. In fact, the code interpreter works like this: you ask a math problem, data analysis problem or a coding issue– something logical, that is. LLMs suck at logic. But they can write good Python code, which when executed… can reach the desired answer. This is what the code interpreter does – allows you to do math, analyze data and more by going through a coding step. For a GPT like ‘The Negotiator’ this is obviously not needed but for a ‘Data scientist’ GPT you would definitely need it.

Based on this, we understand that the GPTs are nothing too special. They are an INTERFACE to simplify or complicate your ChatGPT with custom instructions and other capabilities. Finally, you can also add actions.

6. Actions – You can ask your GPT to perform actions OUTSIDE ChatGPT. For instance, it can ping your API or can send you a Slack message, things like that. This is very interesting in many ways. Its true power? Actions allow your GPT to act as a ‘plugin’. Thanks to the Actions, we now have an easy interface to create plugins or talk to our own apps.

An example

One of the best examples was provided at the end of the keynote by Romain Huet from the OpenAI team.

Watch his example (~5 mins long) from min 33 of the keynote.

Funny story about Romain – I met him at the end of the Dev Day. I didn’t recognize him and I asked: what did you talk about?

Romain: ‘I was a part of the keynote, right after Sam.’

Me: ‘Cool. What did you talk about?’

Romain: ‘It was a full example of how to use GPTs and Assistants API.’

Me: ‘Can you be more specific?’

Romain: ‘I gifted $500 to everyone in the room’ (and he actually built a program live in front of us to send $500 in OpenAI credit to everyone in the room – it was amazing – thanks!)

I said: ‘Ooooh… I remember’.

Truth is.. I remembered the act of giving (for me: getting) $500. But my brain was a total mush by then. It was 10PM in the evening and I had been giving my 120% all day.

On the spot, I couldn’t remember who Romain was or what he talked about.

In retrospect, Romain gave literally THE BEST 5 MINS of the whole conference – succinct, to the point, practical, and super exciting. Watch it.

Romain, if you are reading: Sorry, for the bad taste when we met. Maybe this bad experience will help you remember me!

We can talk about this more but let’s focus on what matters: are GPTs useful?

GPTs are actually genius but I don’t see them being that practical

I must say – I absolutely love the concept of GPTs.

I don’t like the feature because it is oversimplifying a complex prompting task and creating wrong expectations, especially because you cannot see the custom instructions and files provided.

But I love the concept.

Here are 6 reasons why GPTs are genius:

1. GPTs just sound cool. My own GPT, your GPT, my company’s GPT. Rad! I can guarantee you that this is how customers talk about these things. And now they love that they have ‘GPTs’, although they used to have them before, too!

2. Everyone is talking about GPTs. ChatGPT is no longer ‘ChatGPT’ alone. It is now eating the whole ‘GPT’ realm. CoachGPT? This is probably a GPT in ChatGPT, right? Or is it this website:

https://www.coach-gpt.ai/

1. . Not sure… Similarly, our company will probably experience some hit on the name (Team-GPT – is this a GPT where you can do team stuff… or what? No, no. Team-GPT is a different app where teams adopt GPT models - just check us out!!!). By adding this feature, ChatGPT no longer needs to protect the trademark GPT (which they failed to secure). They own the word GPT thanks to GPTs. Kudos!

2. GPTs make custom instructions cool. 90% of the point of GPTs is the popularization of custom instructions. Custom instructions are so important that it was 1 of the only 5 or so features ChatGPT ever shipped in its 1st year of operations. Yet, people didn’t really get it. Through GPTs, you don’t need to know how to use custom instructions. Simply use GPTs of other people who can.

3. GPTs = use cases. For many people, GPTs will be the way to find out about new and more use cases. OpenAI don’t have to educate you on use cases, if a GPT exists for your use case.

4. They are launching a marketplace for GPTs. Anyone in the world can create their own GPT and then share it with the community. This is the best way to popularize them. Give the power to the people and wait for the UGC (user generated content) to flood the market.

5. OpenAI mentioned they’ll be splitting the revenue with the best GPTs, thus disincentivizing people from taking their best GPTs outside of ChatGPT. Who wouldn’t like to split revenue with ChatGPT?

At the same time, here’s why I’m dubious about them.

One of the initial GPTs (featured on their page here) was:

Your tech advisor will help you with… your printer?

Apparently, GPTs will be created by humans. As long as humans don’t understand this feature well, it will suck. And this was created by the OpenAI team. Imagine the rest of the world…

I don’t think we are ready for the pace of AI development we are seeing right now. OpenAI are pushing 100 things out at the same time, and we still haven’t digested the main one (ChatGPT), which came out a year ago.

This is why I don’t believe in GPTs.

They are too genius for their own good!

Not to mention the fact that I’ll create my own ‘Tech Advisor’ and so will 1 million more people. How do you tell the good ones from the bad ones? With ChatGPT? This has to be thought through.

GPTs are brutal (bad) for the ecosystem

Having in mind that this was DevDay (a developer conference), the announcement of GPTs was rather brutal. It didn’t help developers in any way.

On the contrary… with one swift movement, OpenAI killed a million businesses!

Many of the apps that were built on top of ChatGPT, like ‘coach GPT’ or ‘negotiator GPT’ or ‘tech support GPT’ or whatever. They were all… a ChatGPT wrapper + a smart prompt.

A prompt so smart… it deserves to be an app. (as we joked in our company)

GPTs are just this – your ability to build an ‘app’ inside the ChatGPT app, making ChatGPT an app of apps.

If your business was this fragile – it is already dead.

I’m not defending these businesses, I’m just stating the facts.

Assistants API

Since the target audience for this article is NOT technical people, I’ll go quickly through this one.

OpenAI also announced the Assistants API.

This is a new API which will allow developers to build technologies similar to the GPTs we described above. Or even more – agents, and other curious things.

Right now, through the Assistants API you can get the custom instructions, RAG, code interpreter. At DevDay I was told that the idea is to be able to replicate GPTs through the Assistants API 1:1 => the two should have the exact same capabilities.

However, for now this is the situation:

Is the API good? Don’t know yet but our team is very thrilled about the Assistants API. It will solve many, many problems for us.

First of all – Knowledge (RAG).

Knowledge (RAG, aka retrieval-augmented generation)

As mentioned above, RAG refers to retrieval-augmented generation.

It solves 2 problems:

• You can add your own knowledge to the ‘chat’ (simplified), e.g. PDFs, txt files, spreadsheets

• It is the best-known way to limit the hallucinations, so many people have raised money and are working on this problem with the hallucinations

While I was in San Francisco, literally everyone was talking about RAG. Most of the conversations went like this:

Person A: RAG, RAG, RAG.

Person B: RAG, RAG?

Person A: RAG!!!

Person B: RAAAAAAAAAAAAG!

I wish I was joking. But since at Team-GPT our main focus is not RAG, we were looked down upon (a bit)… cause if you don’t do RAG, what are you even doing???

We have developed our own Team-GPT RAG. However, to have a good RAG, we anticipated we would have to buy, not build, so we haven’t invested too much. Thanks to OpenAI, we can now use their RAG, which was developed by the worlds’ leading AI engineers. These people get paid around $0.5M-$10M/year, so this saves us a lot of money.

Assistants API - Fatality

At the same time though, many startups were building RAG systems.

If you ask me – this was a very big blow for the whole startup ecosystem. Hundreds of millions of dollars invested in startups that are building different types of retrieval systems. They all got hit very hard by the OpenAI announcement. All these startups now must pivot, while the VCs have just lost a bunch of money.

Whoever was not killed by the GPTs announcement, truly suffered by this one.

It was interesting how little excitement there was in the room about this announcement. No claps.

A bit later during the day I met John Allard, Engineering Lead, Fine-tuning Product Team at OpenAI. He held the talk Maximizing LLM Performance at OpenAI DevDay.

I asked him: ‘Good job on introducing the RAG! I love it but didn’t you slaughter the whole audience?’

John answered: ‘OpenAI’s RAG is quite simplistic. We are handling the general use cases but if you are looking to sell RAGs to Enterprise, you’d need way more than this. Therefore, folks that are working on RAGs should continue doing so… just don’t make them overly simplistic’.

And fair enough. If you are in the business of RAG, you should invest at least 6 months, 1 year in building it. If you built your RAG with LangChain for 1 day, you are not in the business of RAG.

But the bad taste was there. What if they double down on RAG? No one is safe.

Finally… is the API any good?

The Assistants API was free until yesterday, so we couldn’t benchmark the price (and that really matters when you are integrating a new API). The reviews are not thrilling, though…

Evening reception at the de Young Museum

At 4PM they kicked us out of the conference venue. There was an evening reception at the de Young Museum. I must say – it was really stylish!

10 points for OpenAI!

OpenAI also provided us with $100 in Uber credit, so we can move from one venue to the other. Together with the $500 in API credit, this amounted to $600 in gifts (the ticket to attend cost $450), so that was quite nice of them. Thanks!

I shared an Uber with an incredible Mexican lady (that I already mentioned above).

We shared the ride with another person, too. They were closely affiliated with OpenAI, so I don’t want to name them, but the second thing they said after their name was:

‘I’m here to witness the anthropological effect of OpenAI destroying the businesses of all the people in the room. I’m very interested to see how OpenAI handles it.’

Really? You came all this way for this. You didn’t have to make the effort, pal!

I studied in Western Europe + I sell products mainly in the US. But I’m just a guy from Eastern Europe. In our part of the world, when you smell bullshit, you call it out. Especially if it is an elitist type of douchebag-gery – it really triggers us. This was the only fight I picked during OpenAI DevDay. Strategically, I shouldn’t have – because in business you never know. But morally – I had to see the anthropological effect of using my broken Bulgarian accent to shut the posh guy up with ‘smart words’. Rant over. Sorry – but this is an honest review!

If the person in question is reading – sorry dude but there’s a lesson you needed to learn, hope you did!

Either way – almost everyone else was really extraordinary. I’ll tell you about one other guy I met right after the ‘unpleasant debate’ I had.

I met a guy who is leading Product at FreeStyle Libre. This person has literally changed our lives at home.

My wife has diabetes, and our lives often revolve around regulating her blood sugar. In the last few years, there have been major advancements in the field with devices like FreeStyle Libre and Dexcom that help people with diabetes measure their glucose levels continuously and receive notifications (even on your phone) when it is low.

This is an extremely serious issue for millions of people worldwide.

And it was not only him.

Almost everyone at this conference was as impactful as this guy.

But the guy with the biggest impact was surely Sam Altman, the (former? Future?) CEO of OpenAI.

It was possible to simply go and talk to him. And I did!

Thanks for the memory, Sam!

I’ll finish off with something about you.

After party with Grimes

The day was simply never ending.

At 10PM we moved away from the de Young Museum and into a club downtown.

Inside, I saw some weird signs.

A neural network, with a star, saying ‘COME AND TAKE IT’ was leading you to the dance floor. If that’s not a nerd magnet, don’t know what is!

Looking around, I saw 99% guys. Literally there were no more than 10-15 girls in the club.

What could I expect from an after party in a developer dominated field?

The club had a special guest: Grimes.

Oh, and she was good. She was very good!

Grimes is apparently a DJ, artist, musician and many more. They say she is very smart (I haven’t researched her a lot), and everyone was saying ‘Grimes is really into AI and is very smart’. Was she? Probably. But it really doesn’t matter for the work she was doing that night.

She was DJ-ing and it was magnetic.

I had NEVER seen a better fit for a live performer with their audience.

A hot girl, who is apparently smart and into AI, playing some good music in front of 99% guys.

Everyone had a good time. Software engineers dancing like crazy? That was new for me! Many attendees I talked to said this was ‘the best performance they’ve been to’.

Others who had seen Grimes before were thrilled to see her again.

Entranced AI engineers vibing with Grimes… I loved it!

And OpenAI paid for everything!

There was an open bar, there was Grimes and probably more. Interestingly, Grimes was also invited to OpenAI DevDay. There were some theories that this was to make Elon Musk angry but I don’t think anyone cared about him that day.

It was the day of OpenAI.

And to top this off – I learned something new.

Remember the ‘Come and take it’ sign?

There were more.

And more.

These were placed by a movement I didn’t know about before.

Effective accelerationism.

I asked around… but no one wanted to talk about it.

But everyone knew.

e/acc.

These are the things you can learn only when you go there.

Summary

OpenAI DevDay was far from the best conference I’ve been to, but the event was rather nice.

I’d give it 8/10.

Regarding OpenAI and the future of ChatGPT?

I was going to give it 10/10.

It was the most solid keynote announcement I have ever seen (after Steve Jobs announcing the iPhone).

And no wonder – Sam Altman and Greg Brockman started fundraising literally the same day. This was their plan all along. Greg met with Macron (President of France) and Sam met with important people, too.

But the board of directors didn’t like something… was it the keynote? Was it the fundraising? Was it the ‘acceleration’? So 11 days later, on Nov 17, they fired Sam Altman.

I was at DevDay. Talked to Sam, saw the keynote, understood the stuff he said. I’ve also been in the industry (both AI and startups) for 8 years and this move is unexplainable to me.

As I’m writing this, Sam Altman is no longer CEO, Mira Murati (CTO) is now interim CEO.

It is hard to unpack this, and I don’t think this is the right place but I want to say this:

• On 6 Nov 2023 (OpenAI DevDay), OpenAI seemed like the biggest company that will ever exist

• On 17 Nov 2023 (after firing Sam Altman), OpenAI seemed like an inevitable crash and burn

The specifics are yet to be made available to outsiders (and insiders lol) but one thing is for sure – the next few days will determine the future of AI.

From the looks of it, Sam Altman will either:

• Rejoin OpenAI as CEO, change the board of directors and become stronger than ever

• Start a new venture (probably with Greg Brockman), raise a lot of money (investors love him), poach a lot of talent from OpenAI (employees love him), and outperform OpenAI

When I posted my photo with Sam, someone DMed me:

‘You are in the presence of royalty. Sam Altman is the new prince of tech’.

Sam Altman was dethroned as the prince of tech.

But he might return as the king.

🖖

P.S. Latest news are that Sam Altman and Greg Brockman and joining Microsoft in an ‘new advanced AI research team’. Read more here.

P.P.S. Sam Altman is back as CEO of OpenAI. Greg Brockman is also back. The board of directors of OpenAI is now changed with the new one being more supportive of Sam Altman and his vision for OpenAI.

Revenue: $80,000
Net revenue: $ 47,819.58
Time: 170 days
Users: 9597

Here’s some eye candy before we get into the real article.

And the other one:

9198 users 4 days ago. 9597 as I started writing this. 9611 as I’m reading the article for the last time. At the time of publishing (it was scheduled for the next day) 9702.

The business is going well, thanks for asking!

But...

I want to talk about something else: Getting ready to scale.

I think it is all about providing as much information as possible to others, so they can make better decisions about and around Team-GPT as a project. The more they know, the better informed decisions they’ll make.

But I have no time for this.

Time has become my absolute enemy.

As the number of people in the project is growing it is becoming impossible for me to communicate with every team member. Some people I still haven’t met.

It’s quite sad and I don’t want it to be that way... but there is no other way.

The team itself is quite solid and I trust that I’m not always needed.

This article is not about ‘How we got 10K’ users.

It’s about... a startup CEO, trying to communicate better, as Team-GPT is preparing to scale.

Why?

As the CEO, by default, I’m the owner everything.10K users is quite a milestone and I must become more and more strategic.

I’ve learned that my job is to:

• tackle the issue with highest priority

• find a new owner for it

• and move forward

So far:

✅ MVP + Validation
✅ Go-to market strategy
✅ Financing
✅ Dev
✅ Product

Finding a new owner for a big topic means: hiring the right people.

I know very well that:

• 1st time founders focus on product;

• 2nd time founders on distribution;

• 3rd time founders on the team.

I’ve been a founder 3 times already. I’ve made most mistakes, so I get it. Truth is, you have to focus on all of three. And in the span of Team-GPT as a company, I’ve focused on product and distribution simultaneously for the first several months. Now I’m almost out of product.

My current challenges are distribution and team. I’m tackling them together and bit by bit, the right people are falling in to the right places.

One of my co-founders (Yavor) often likes to ask me:

What is your focus now?

An amazing question that every team member should ask themselves.

My focus now is:

👉 Marketing - our marketing is amazing but it is not scalable. We don’t have a single full-time marketer in house. This is really bad. But I need to find the right person. Currently, we are growth hacking the s*it out of this. 10K users without a marketing hire.

👉 Education - our users need to be educated + my comparative advantage over the competition is precisely education. This will be huge for us.

👉 Sales - I’m already focused on the Enterprise sales. I need to improve and I’m reading a lot on how to do this properly + getting mentorship. I have some experience but this is by far my weakest skill.

And here come some of the things we’ve been doing to prepare to scale and solve many of the problems that we’ve faced or will face.

1. Onboarding videos

See Loom videos HERE

Videos: 22

Duration: ~ 4 hours

On 3 July 23, I sat for about 2 days, recording myself explaining in depth what Team-GPT is and what are we going to do with it.

I expect every new team member at Team-GPT to watch them.

All team members need to know the history and the planning, so we can reach the future.

Some videos are 1 min long, others are 20 mins long.

To this day, I send the onboarding videos to every new person coming to the project. No matter if employee, investors, or else.

Those who watch the videos make the initial cut to join the team.

These videos literally contain all our ‘secret strategies’ about the future. A gold mine for competitors.

But they don’t have our team.

2. Building in public on LinkedIn

Follow me HERE ... or click here!

Step number 1 of the journey was building in public on LinkedIn. This got us at least $4000-5000 in revenue. I know LinkedIn is really helping sell more.

One of my co-founders (Ilko) suggested the #buildingpublic and I’ve been doing it ever since.

I quite like it. And it is very useful to me.

I’ve had to communicate so many things to so many people, that I often just write in public channels and then copy paste to LinkedIn.

Like this one big article.

I’m happy to update anyone on the Internet: users, investors, partners, enterprise clients, or just.. friends.

Plus, I barely have any secrets nowadays. I don’t have time to think about ‘who knows about what’. No time for drama, mama.

Being 100% transparent and genuine at all times is really helping me build trust.

Even if I get into conflict with someone, it’s helping us move faster.

A bit of conflict is never bad because it gets real ‘words and emotions’ out. You face them and you go through the problems.

3. Recording myself building the ‘ChatGPT for Work’ course

See YouTube playlist HERE

I’m preparing a course on ChatGPT for Work.

I’ve spent 6 months gathering information and learning how to use ChatGPT properly. I’m ready to teach it.

And I know I’m damn good at it - I’ve got 1.2M students on Udemy.

Apart from this, many people have asked me how to create a good course.

How to create a course + Build in Public + Transparency + ChatGPT =

I decided to stream the whole thing.

(weird I know)

SEE HERE - How to create a course about ChatGPT with ChatGPT.

I’ve recorded 15-20 hours of video content filming myself while thinking about the course ‘ChatGPT for Work’ and coming up with ideas about it.

Why?

I have to transfer this information and knowledge. For others who want to create a course. And for my team. Some of them want to know as much as possible because it helps them do better work (being in sync with me)!

Best part?

I’m using Team-GPT while doing it. This is ‘practice what you preach’ 101.

And it works, people! I’m 20x more productive. ChatGPT is truly amazing.

Actually many people have watched bits and piece of it. It is so unfiltered that I actually recommend it.

Deep within the lectures, I leave ‘Easter eggs’ for the true fans.

My wife said: ‘some day you will regret doing this’.

4. Phone calls

This actually works quite well for me.

I call people on their cell phone and it helps us sync. It is not old-fashioned, it is pretty effective. I highly recommend it.

Typing has become too slow for me to convey information.

I normally walk Lucky, the best dog, while syncing with different people.

Works quite well for my stress as well.

Walking is actually super beneficial for people who have high stress and/or work on a computer. 45-min walking with the same tempo gets the who cardio-vascular system to work and helps recover from all the stress and strain.

Highly recommended.

On average, I walk for about 11,000 steps every day.

That’s the time to sync with my team. Walk and talk :)

Heard it from Tim Ferris.

Tim, if you are reading this, I would love to meet you some day.

5. Meetings schedule

I’m scheduling all my meetings in these 3 days: Mon, Tue, Wed. On these weekdays, I have opened for anyone and everyone in the world.

I’m taking all the sales calls with Enterprise clients, so I need to be available. At least in the beginning I’m probably the only person who can sell the Enterprise software. Head of Sales is a highly needed role and I hope to soon have a solution for this. But still... in the beginning FOUNDERS have to be the sales people. I must do it. Good thing is, I’m receiving a lot of support from Jeko, a friend of mine, who has negotiated deals for $60M. Our sales super pro.

Thursdays and Fridays are blocked in my calendar.

I reserve these for important LIVE meetings, where I need to travel across the city or do something out of the ordinary. I try to use this time for deep work. I’ve started booking deep work inside my calendar.

I’m very productive when working on my own and I need to have this time, so I can create more.

I don’t want to become a CEO who never touches actual work.

I’m a maker. I like implementing stuff.

It’s where the magic happens.

Paul Graham, founder of YC, said (and I’m butchering this, he said it way better): people who are makers should not have any meetings on some of the days. The problem is that by just ‘knowing’ you have a meeting you are stressing out and cannot chill to start creating and making. Here’s how Paul Graham put it in Makers vs Managers.

6. Documentation

Needless to say, almost every call is recorded so others can watch it if needed.

Also Notion is heavily used. We create a lot of our internal documentation and processes there. OneNote is also a favorite of mine - all my meeting notes are there.

Most of the tasks are in Trello and we try to document there as much as possible, too.

All information is open to everyone, as long as they want to read it.

You need help with something? Me or someone on the team will record a Loom for you. Then we will upload it on YouTube.

That’s how our knowledge base is getting built.

One Loom at a time!

P.S. Then we take the Loom transcription, feed to Team-GPT and write the prompt ‘Take this video transcript and create a how-to style guide with steps 1 to X, based on the information shown in the video.’. Always repurpose content in as many ways you can think of. You’ve already created it, time to reuse it.

7. Async Looms

I’d often record Looms at weird times and send them to teammates. This is saving me so many meetings.

So many words are spared, you can’t imagine!

I say it once, whenever I have the time and then others can see it, whenever they have the time.

It’s a fascinatingly simple concept, which has changed my life.

Definitely use Loom.

8. Slack messages

I think Slack messages are underrated.

Writing in public Slack channels to update people is super important and everyone should do it.

Give your teammates information and context and you’ll do better stuff together.

DON’T just chat on Slack.

Write actual informative texts.

9. Reuse, recycle, reduce. Repurpose

If you can create texts, videos, and other stuff that can be reused, do it. In marketing we call this ‘repurposing content’.

Create content that can be shared with multiple counterparties:

• Record a Loom video for a teammate or client? Don’t say their names or anything personalized and just upload on YouTube later or send to others.

• Showing someone how to do something on a Zoom call? Record the meeting and add to the team knowledge base

• Have a good chat with ChatGPT? Use Team-GPT to build your library with good chats and prompts (while collaborating with your team). Then let others reuse it. Okay. I’m selling software here, sorry. But really, this whole article is, anyways. Jokes aside, Team-GPT is actually an amazing product and deserves to be taken seriously by you. It’s free for small teams.

• You write a Slack Status Update for everyone... post it on LinkedIn

That’s what I’m about to do too. Here’s my latest Slack messages to the team.

Some numbers that I’m looking into

10. MRR - Our monthly recurring revenue (MRR) is growing at 66%

Just 30 days ago we were making $2100/m and right now we make $3500/m. Note that this is till not much, however, the numbers are growing steadily.

11. AppSumo sales

$27,500 so far (for us) out of which $6747.30 in September alone.

More than 50% of our sales so far have come from AppSumo.

These sales are in AppSumo and they are our LTDs.

People from AppSumo are early adopters and we want as many of them to use the product as possible. This is our initial user base and they battle-test the system every day.

Thanks to AppSumo people, we know when we have bugs, when system is down, etc.

Most recently, we have 3 Enterprise clients who are interested in us, who learned about us from AppSumo.

This was the plan all along to convert them from B2b to Enterprise, where the real money is.

12. Total users

Our total number of users have grown to 9597.

This is very close to 10,000 and this will be a huge milestone.

10K users is serious shit! Just think about how many apps of people you know have reached 10K users.

Once we have this many users, we can start thinking about community building, referral programs and all other types of stuff.

This can help us have a good Product Hunt campaign, launch a mini product better and much more.

We are in their inbox and can communicate directly to them + they usually love the product.

13. SEO

Our SEO is extremely good. We have been adding most of the people to our system through AppSumo or SEO.

Our initial efforts from April-May have paid off tremendously and people are actually finding us on Google a lot.

We’ve got a total of 18.3K clicks and 329K and our organic traffic is growing steadily.

There is nothing better than organic growth and organic ranking - it pays in the long run 1000x.

In one of the images you can see a list for our best keywords. But we don’t only rank for them.

We rank for 690+ different keywords and our DR (domain rating) is 15. This is still not high but out of the ‘scammer zone’.

We will improve this with time and this will help our SEO even further. Most of the words we rank for are in the US and GB which is ideal!

Our SEO team is still struggling a bit and getting behind but I think we are almost there. They’ve published 4 new articles and will be publishing another 12 by the end of October.

Once we get a marketing hire (I hope we found 1!), we will also push more content onto the website.

The more text and information we have - the better.

14. Website traffic

Here is our website traffic overall.

75,000 users.

Imagine 75K people know something about us.

Once we can focus more resources on marketing, we will start retargeting these (we already are doing it but not very aggressively).

We will retarget them with hundreds of videos that we are currently publishing all over social media. And they’ll remember us.

Many of them are actually using the product itself.

Our overall analytics tracking is not amazing. We don’t have the proper person in place to fix this and I don’t have time.

We improve this area.

15. Accepted in Paddle AI Launchpad, Microsoft for Startups, and AWS Build Accelerator

Everything is so dynamic that while I’m writing this update at ~11.29PM, my co-founder Ilko is joining with another piece of news. We are accepted in AWS Build Accelerator.

We are already in Microsoft for Startups ($5K credits) and Paddle AI Launchpad ($25K cash prize).

Every bit of credits or money we can get our hands on is needed.

We are bootstrapping and have to be as efficient as possible!

16. Product analytics tracking

~2300 active users

In terms of product analytics, we are very good. We are using a service called ‘June.so‘ and we have some pretty good data on what’s going on in the app.

Every day we are hitting higher and higher usage.

Daily active users are ~1000 right now.

But the proper metric to look at is ‘Weekly active users’. Most real users actually use ChatGPT 2-3 times per week. Therefore, daily usage is not so relevant.

2300 active users is some very good engagement and we should be thankful that so many people are finding the product useful.

ChatGPT has so many more features than us - images, speech, pdfs, whatever.

And yet these 2300 people prefer to use Team-GPT because it solves a much bigger problem for them: collaboration and knowledge sharing.

17. Social media!

Finally, some reporting on our social media.

We’ve created 100 social media videos with Simona (Head of Product). We are posting 1 each day for 4 months.

CLICK HERE -to check them out

I’m sending the report which I sent to the team which is dealing with the social media stuff.

I’m just copy-pasting, so there are also comments for them also some tasks (for them ). It’s in Bulgarian because we just speak in Bulgarian with them.

(don’t worry guys, I used ChatGPT to translate, haven’t fact checked 100% but ... no time, remember)

Looking at the success of the videos so far, I notice the following.

• Instagram

Almost all videos have around 50-70 views. Exceptionally few. We need to change something.

Probably Kali was right that we shouldn’t spam there. I think she suggested we post 2-3 a week to let the algorithm show them.

Something like that, right? Please try whatever you deem appropriate on Instagram to get more visibility.

Once you’ve figured it out, write it down here (or even better in a Trello card) so we know when we made the change and can track the results of the experiment.

• Twitter

We aren’t having much success there either. However, I think we don’t need to change anything. Let’s focus on the other channels, as they have more potential.

• Facebook

Facebook seems to be going well in my opinion. The Reels have around 200-500 views on average. Almost no videos without views. That’s a success for me. 100 videos, averaging 350 views, are 35,000 views.

Good enough. Let’s not make any changes there for now either.

• LinkedIn

I post these videos on Saturdays and Sundays. People really enjoy them and it’s not bad, but I try to post less often so as not to spam them. There it’s not from the company account and it feels a bit strange.

Overall, it’s good content that complements the things I usually write.

I will experiment there and if there’s anything more interesting, I’ll let you know.

• YouTube

Things are going very well on YouTube.

We already have about 200 subscribers, 14K views and a generally stable viewership of around 200-300 views per day.

We aren’t mega viral, but we’re mega stable, which is important. The subscriber count is growing every day.

Generally, we’ll be posting more and more things there.

-- Kali had the idea to post Shorts, which I rejected before. It was very important for me to have horizontal videos. And it’s still important to have them - that’s why Tsanko makes both horizontal and vertical ones (thank you!).

Despite that, I’m open to trying out Shorts and see if there will be any success. I’m attaching an image of our most popular videos.

Our Massive Mistake - TikTok

TikTok is the network with the highest potential. There we have videos with varying success, 300-500 views. One of them has 7,000 views.

The problem is that our content is perfect for TikTok, yet we’re not handling it well. We recently discovered that our TikTok is based in Bulgaria. Because we created the account from Bulgaria.

This is a rookie mistake, but “what can we do” overall - no one realized. Right now, I’ve started the process to “create an account in the US” or something :D

In the worst case, next month when I go to the US, I’ll do it. The scheme is for the account to be created in the US and to release all videos to an American audience (or at least some English speakers”), not in Bulgaria.

In short: We will create a new TikTok account and start from scratch.

Sorry folks, but better a month late than never to realize.

• Future social media

Just a little info for the future: in 2-3 weeks we want to start running ads. To boost some posts and primarily to retarget the people who have visited our site with these videos.

I’m thinking of starting with the Meta ecosystem and then YouTube, somewhere around the beginning of November.

It’s a wrap

Yo! 🖖

Abrupt ending.

Sorry for the long read.

Don’t know if I’ll write another.

But if you finished reading this, drop me a line on LinkedIn.

That’s how I know you are up to speed with Team-GPT.

It could be useful to know.

The problem

Netflix and The Queen’s Gambit have taken the world by storm. As a result, everything chess-related exploded. Chess.com reported about 300% growth in their monthly sign-ups after the show aired.

It seems like many people want to play chess. I even got my chess board out of the cupboard and started teaching my wife.

Problem is – chess is not an easy game to learn. But it is worth committing time as it is a great way to learn the importance of responsibility and discipline. Similar to life, each tiny decision, has its consequences later on.

Even with the exploding number of free online materials, they still cater to the very beginners in the game. In fact, there are ‘only a few chess coaches ready to drill into your game, helping you to become better’, explains chess coach Bernard Chan (P.S. you should check out his newsletter).

The solution

What if there was someone who used to be the school nerd?

Someone who spent days and days mastering chess… before it was cool.

They would definitely possess an in-demand skill they could monetize, wouldn’t they?

Such a person could definitely devote several hours per week to provide online chess lessons to someone looking to improve their game.

Were you that shy person who preferred chess to hide-and-seek?

Target customer

To be honest, just about anyone interested in chess.

Demographically, that would mean there is already some information for you to go with – about 22% of Chess.com players are women and 78% are men.

Interestingly, after Queen’s Gambit, those numbers became 27% and 73%. Although there is no need to discriminate, it sounds like 3 out of every 4 of your prospective students will be male.

How to start this as a side hustle

Identifying the need for your services is the most important step.

Done?

There are several more:

1. Prepare the lesson materials

2. Finding an online platform where you can do that

3. Finding people who would pay for your services

Prepare the lesson materials

There are different ways in which you can approach this. However, it seems many chess coaches struggle in the beginning. If you ask me, I’d recommend the system which my first chess coach used.

He opened the board put two rooks (from the same color) and the two kings and asked me to check-mate him.

Then he took a queen and the two kings and asked me to do it again.

Several lessons later we started exploring different openings and so on.

You surely know how to get around this.

However, one bright idea is to go to Chess.com and check out the Lessons tab. It will give you many ideas on what some of the best lesson approaches are.

As chess instructor David Siamon suggested, you also can go through Master games with your student (which is greatly supported by Lichess but also by chess.com, with one prominent example here). However, this may be more useful for your more advanced clients.

Finally, be sure to also spend some time on 1-move and 2-move checkmates or Queen sacrifice. All of these are extremely useful as the more the student plays, the more likely it is for them to spots them during a game and win. In this way, they’ll not only be better but also directly appreciative of your coaching.

If you don’t want to build your own curriculum, it is worth going down the beaten path. Great places to start are:

• Comprehensive Chess Course Volumes I & II by Roman Pelts. Here’s the book on Amazon.

• Learning Chess Step by Step by Rob Brunia, and Cor van Wijgerden. This one is available online over here.

• How to Reassess Your Chess: Chess Mastery Through Chess Imbalances by Jeremy Silman is a student favorite. You can read it not so much for new knowledge but for techniques you can then transfer in your teaching process.

Finding an online platform where you can do that

Finding an online platform is easy. In fact, you can check this box already. I guarantee you that 3veta.com is a great place to start. You can host a video chess session there and charge for the lesson. Moreover, you can also create your own booking page. I created my own in two minutes or so: iliya.3veta.com. It’s not amazing, but showcases who I am and what I can help with.

Finding people who would pay for your services

The third piece of the puzzle is marketing.

This becomes a bit easier once you have your own website. You can try sharing it in different places as this is your ‘business card’. The best place to start is from your friends and friends of friends. Post on social media and see if anyone is interested.

In fact, why don’t you post your whole games on Instagram (or even TikTok?). Coach Chad’s Instagram is a great example of how you can seamlessly grow your following by posting your games.

Anyhow, it is fundamental that you understand that no one else is going to lead generation for you. Therefore, you should be looking into finding people and growing your personal chess instructor brand. While social media are kind of clear, it is hard to find quality leads there (e.g. interested parties who would be willing to pay for a prolonged period of time).

Forums and communities

Your next stop is chess-oriented forums. Online chess servers (chess.com and lichess.org) definitely have chess forums. This is where chess happens, so you will want to be there and interact with people.

Where you have forums, you have social media communities. The biggest one on Facebook is the ‘Chess Teachers and Learners’ group.

In terms of live communities, where do the people interested in chess go? Well, quite a lot participate in chess tournaments. Tournaments are a great place to meet people who are interested in improving their game.

Live streaming and video recording

While forums have always been the go-to place to discuss, the recent moves towards more and more video content point us to video. Since chess is a game, after all, you can deduce that apart from YouTube the go-to place is: Twitch (video + gaming).

There are many creators popularizing chess on Twitch. Some of the most popular chess streamers you can check out are GMHikaru, GMNaroditsky, and chessbrah. There are also many women, who are working on including more women and girls in the game over there.

If being on camera doesn’t make you feel anxious, Twitch is definitely a place to try out.

Finally, whatever you create on Twitch could be uploaded right away on YouTube, helping to grow your presence there.

Schools and local communities

Go to the local school and ask if there is a chess club. Try to promote yourself there – every parent would be happier if their child played chess, rather than Fortnite. Moreover, with the development of the COVID-19 pandemic, many older coaches in schools around the US refuse to continue coaching in the schools.

Furthermore, these people may are usually not as tech-savvy to bring this online. While it is understandable for the coaches and disappointing for the schools, it poses a great opportunity for younger chess instructors who are looking to grow their practice.

Finally, even if you’ve got chess clubs around you, make sure to visit them at least once. These are still one of the places where real-life chess happens. Over there you can find both prospective students, as well as fellow chess coaches who you can exchange ideas with.

Group classes

Speaking of schools, children tend to prefer group classes to 1-to-1 sessions. The drivers of this are the preference of kids to communicate and play with other kids, as well as the financial side which falls on their parents.

Creating group classes is a great way to get a couple of students together and bring the price of your service down. This will save you time while making the experience more immersive for them.

Apart from these obvious benefits, there is something much more important from a didactic point of view. Group lessons allow students to play real games with each other. They play different styles and with varying proficiency of the opponents.

Finally, you can set up small tournaments between them online. They simply need to screen share each game and you can simultaneously watch overall games at the same time.

Speaking of mock tournaments - this is the best time to ask the students to record each of their moves. This will get them used to the chess notation and prepare them for their chess games later on. Once having recorded the moves the game is documented and can be reexamined. It is a great practice to go through games together with the instructor. Then each student can explain their thoughts and motivations for each move, while the teacher can point out mistakes and blunders.

Homework

Your chess lesson may be over but this should not stop the learning for the student. Make sure to give them homework. This could be in the form of reading resources or watching a video. However, why not make it fun for the students?

1. Chess puzzles. You will surely have the most success with chess puzzles. Chess puzzles are an incredible way to train the eye of the player. Once they have solved enough of these, they will start seeing the same positions in real games. Many chess instructors confess that chess puzzles have been not only quite fun but also extremely useful to stay on top of their own chess game.

2. Play X chess matches. You are teaching your student(s) to be better at chess. Well, ask them to play it more often. It is a good idea to set a number of chess games they must play before you meet again. Ask your students to record the games (some platforms do this automatically). In this way, you can analyze these games together in your next lesson.

3. Play a certain opening exclusively. Say you’ve just discussed the Reti opening during your last chess lesson. For homework, ask your student(s) to play only Reti for a full week.

Don't get discouraged

While many people want to learn chess, most beginners get discouraged quickly.

To reiterate the beginning of the article: chess is not an easy game to learn.

In turn, you (as a chess coach) should not get discouraged either. Having a high dropout rate is just a part of the job.

And keep in mind, they don’t always get frustrated with the lack of progress. Sometimes they simply reach their goals and need some different coaching from a different coach.

But as long as you love what you do, teaching chess won’t feel like work at all.

Tournaments

If neither you, nor your students got discouraged, then it is time to get serious. There are various tournaments for all levels of chess proficiency. As a chess coach, you have probably attended quite a few.

Chess instructor Danny Angermeier shares that the whole environment of in-person chess tournaments is quite thrilling. When you are playing, you are trying to always bring your A game. But that’s not all. Tournaments also allow you to observe players who are better than you.

Many games start in a similar manner. To bring this to the extreme, a student’s game often starts in the same way as that of a GM. On tournaments students can be there live, seeing games which resemble their games but develop quite differently. And in this environment they can grow a great deal.

Extra thoughts

• You don’t need to be the best chess player to be a great chess coach. What is important is to be able to identify weaknesses and be willing to share and explain them to the students.

• Get listed as a coach on Chess.com Coaches section or Lichess Coaches section. Both are a great way to get some more exposure.

• While on this topic, general-purpose posting websites like Kijiji and Craigslist are always worth posting on as you can never get too much exposure.

• You can play chess on chess platforms and offer help to your opponents. Not sure if that’s in sync with Chess.com’s terms of service, so you’ll need to check that out.

• Try offering the first chess lesson for free. This will showcase your skill to your prospective students and make it easier for them to start paying.

• There are numerous free tournaments where people take part. One of them wins the tournament, the rest could be highly interested in elevating their chess game. Attend these events and promote yourself there.

• If you are looking into teaching children below the age of 13, there are some applicable regulations (e.g. COPPA) when doing it on your own website. Therefore, it may be a good idea to use a website like ChessKid and leave them to handle the compliance. Note also that teaching children is not only about chess but also accountability, discipline, and delayed gratification. In the beginning a child will lose hundreds of games of chess with little progress and you have to keep them motivated!

• Make sure to research or revise the most common openings. You should be sure to know what the ‘Queens’s Gambit’ is as that’s one of the more common questions nowadays.

• Sometimes both you and your student may be too tired to learn new things. However, to keep this in the coaching spirit, it can be a great idea to play some variation of chess. I often play Xiangqi (Chinese chess) but you can also try Bughouse chess or Vegas Chess. It is quite refreshing after long hours of concentrated chess.

• Bobby Fischer, Gary Kasparov, Magnus Carlsen, etc. – these guys are celebrities due to more than one reason. Research their stories and tell them to your students – they will love it!

Do you think this is a good side hustle idea for extra income?

I’m sure I’m missing something, so do let me know in the comments 😊

Subscribe now

Dear reader,

This post isn’t designed to blow your mind. It’s designed to do something smaller - and maybe more useful.

It’s meant to remind you of a painfully obvious truth: to get anything done, you have to start.

That’s it. No productivity app, no secret framework. Just - start.

That’s exactly what I’m doing here: starting.

How Writing Actually Happens

Let’s be honest about how ‘writing an article’ usually looks.

On paper, it’s simple:

1. Decide on a topic.

2. Write it.

3. Publish it.

Sure, some posts are just ‘brain dumps’ (as many you’ll read here).

But if you want to write meaningful stuff it’s often messier.

There’s research. Then more research. Notes. Outlines. Keyword spreadsheets? Structure decisions. Length debates. Editing. And then finally - if you’re lucky - words. And then. Rereading. Rewriting.

We turn a three-step process into a 17-step ritual. Then we wonder why we never finish.

The Hidden Step Zero

The trick? Don’t worry about the finish line. Just take the smallest possible step toward motion.

For writing, that means:

• Jot down some notes.

• Open a Word file.

• Type something - anything.

You can’t write a sentence until there’s a blinking cursor in front of you.

That single action - opening the file - isn’t trivial. It’s a mental switch. You’re telling your brain, it’s writing time.

Maybe AI can help. But writing is not about using AI. It’s about using your brain.

The Psychology of Momentum

Here’s the upside of starting: once you’ve written 100 words, you’ll want to write another 100.

Momentum is contagious. Your thoughts begin to self-organize. You find flow. And if you stop? Great. Because tomorrow, when you return, your brain will already be tuned to the same frequency.

That’s not self-help fluff - it’s neuroscience. The brain rewards continuity. You’re five times (okay, I made that number up) more likely to finish once you’ve begun.

Just start

This works far beyond writing.

• Need to edit photos? Open the editor, drag one image in.

• Need to make a playlist? Open Spotify, add a single song.

• Building a financial model? Duplicate an old sheet, rename it, plug in a few cells.

You’re not just clicking apps - you’re building context. Choosing the tool, naming the file, entering the mindset. Those micro-actions are the launch pad.

Starting is a design decision. It’s how humans hack inertia.

This article is the proof

I literally just started my blog.

Opened the editor and merged all three steps - decide, write, publish - into one.

It became a self-fulfilling prophecy.

First a post.

Then a blog.

Have you started? You’ve already won half the battle.

Subscribe now

How about you start being a subscriber? 👆

Iliya