LettsGroup VentureFactory

Someone asked me this recently and I gave a tidy answer that I did not entirely believe. So here is the longer and more honest version.

There are three of us. Between us we have over thirty years of building companies, across three nations. One is an English serial entrepreneur. One is a Swedish product and technology builder, our CTO, who built one of Sweden's largest systems integrators, which was acquired by Cambridge Technology Partners, then went on to build one of the first corporate portal vendors and one of the earliest cloud companies, before most people had a working definition of cloud computing. The third built the platform you are reading about and runs our AI. Before this he built one of the first smartphone operating systems, back when nobody was certain smartphones were going to be a category at all, and it was later sold to Fujitsu. Earlier still he created one of the first cryptocurrencies, an early attempt at a genuinely web native currency, years before any of that became a mainstream conversation.

I mention all of that for one reason, and it is not credentials. It is because building this company was still hard. Genuinely, embarrassingly hard. We made decisions that were fine in isolation and wrong in sequence. We lost weeks to questions that had known answers. We rebuilt things we should have validated first. And at some point one of us said out loud the thing that became the company: if it is this hard for us, what is it like for someone doing it the first time?

The thing we could not stop thinking about

Y Combinator accepts around two percent of applicants. Set aside whether that number is exactly right this cycle. The shape of it is the point. A very small group of founders gets access to a structured system, a framework for what to decide next, and a room full of people who have already made the mistake you are about to make. Everyone else makes hundreds of business critical decisions with none of that.

The uncomfortable part is that the knowledge is not secret. It genuinely is not. Almost everything a good accelerator teaches you exists in public, scattered across a thousand blog posts and books and podcast episodes. The problem was never that the knowledge did not exist. The problem was always access, sequencing and context. Knowing which piece of advice applies to you, at your stage, given what you already decided last month.

That is the gap we kept circling. And the reason we eventually built something is that large language models changed what was possible about it. Not because AI writes decent copy, which is the least interesting thing about it. Because a system can now hold your context, apply a structure to it, and tell you what to do next in your specific situation rather than in general.

What we actually built

VentureFactory is a venture building platform with a seven stage framework called Innov@te sitting underneath it, running from first idea through to exit. Alongside it you get AI co-founder agents covering strategy, product, growth, finance, legal and fundraising.

The part we care most about, and the part that took longest, is that the agents share context. Your finance agent knows what your product roadmap says. Your fundraising agent knows what you recorded three weeks ago. Everything the platform produces lands in your own data room, tagged to the stage it came from, so you are accumulating a record of the company rather than a folder of disconnected AI output.

That design choice came directly from our own frustration. We were all using AI heavily and all quietly doing the same unpaid job: carrying context between tools. Re explaining the business to a fresh chat window. Being the integration layer. The individual tools were not the problem. The missing layer between them was.

Where we actually are

We only launched in the spring, so this is early and I am not going to pretend otherwise. We are currently adding around one and a half customers a day. That is a real number and it is also a small one. I am putting it here rather than talking about momentum, because Indie Hackers punishes vagueness far more than it punishes being early, and because a rate is more honest than a total when you have only been live a few months.

What I take seriously is not the number itself but what sits underneath it. Most people who register actually go on to start building something rather than looking around and leaving. That is the only early signal I trust, and it tells us the problem is real even when the revenue is not yet meaningful.

Why we keep going

Because the thing we suspected turned out to be true. The founders using this are not short of intelligence or drive. They are short of structure. They know roughly what to do and have no reliable way to know what to do next, in what order, given everything else that is already true about their company.

We are also, unavoidably, our own test case. We are three people building a platform that claims a small team can operate like a much larger one. If that claim is false, we will be the first to find out, and publicly. That is uncomfortable and it is also the most useful accountability structure we have ever had.

If you are building something and any of this sounds familiar, you can create a free account at letts.group and tell me where it falls short. Critical feedback is more useful to us than encouragement right now.

The great promise of AI was that one person would soon be able to run an entire company. Nobody mentioned that the company would consist mainly of browser tabs. The solo founder of 2026 has no employees, negligible overhead and a laptop that sounds like it is preparing for takeoff.

It begins innocently enough. ChatGPT helps with some research. Then you discover a better research tool, something clever for coding and an AI that makes presentations without requiring you to spend Sunday evening nudging boxes three pixels to the left. Six months later you have 14 subscriptions and have forgotten what three of them do.  This is called productivity.

The 37-Tab Founder Frenzy

 The strange thing is that most of these tools are genuinely excellent. The research is better, the code arrives faster and things that once required a small team can now be done before lunch. Unfortunately, none of your artificial colleagues appears to have met.  So the founder acquires a new job. You spend Monday morning explaining to one AI what another AI discovered on Friday. By lunchtime you're copying a product decision somewhere else because the marketing AI is still enthusiastically promoting the feature you've just killed. At 3pm the fundraising AI resurrects it in the investor deck.

Nobody has employed a middle manager, which is excellent news, because you have become one.  This is the slightly ridiculous state of the AI native startup. We have developed machines capable of writing software and reasoning over enormous quantities of information, then arranged them so that a human being has to wander between them carrying messages. Somewhere, the inventor of the internal memo is feeling vindicated.  The underlying mistake is treating a startup as a collection of jobs. It isn't. A customer saying something unexpectedly rude about your product on Tuesday can change what gets built on Wednesday and make Friday's revenue forecast look faintly embarrassing. Companies are chains of consequences disguised as departments.

That was manageable when execution was expensive. You had a marketing department because marketing required people, and a finance department because accountants became surprisingly territorial if you tried to do their work yourself. AI changes that bargain. Yet our first instinct has been to recreate the old company in miniature and populate it with agents who never go to lunch but also never seem to talk to each other.  Hence the emerging fashion for the AI employee. Soon every founder will apparently have an AI CFO, AI CMO and AI Head of Strategy. At the current rate, someone will launch an AI Chief Happiness Officer by Christmas and it will spend its days asking the other agents to complete an engagement survey.

There is a more interesting possibility. Instead of making every task intelligent, you make it harder for the company to be stupid. That means the useful bit isn't simply memory. Remembering that you decided something six months ago is nice, but remembering why you decided it becomes considerably more useful when someone proposes the same bad idea again wearing a different hat. A company accumulates arguments, experiments, awkward customer conversations and expensive mistakes. Most software treats these as debris.  But they may actually be the valuable bit. The models themselves will keep changing and today's miraculous one will eventually be spoken about with the affection currently reserved for fax machines. The company's accumulated context, meanwhile, becomes peculiar to the company.

Startup Dreams, Mortgage Realities

 That's the thinking behind what we're building with LettsGroup VentureFactory. The interesting question for us isn't how many AI agents a founder can command before developing the management problems they were supposed to eliminate. It's whether the venture itself can become the shared system, so what is learned in one part of the company doesn't have to be manually smuggled into another.  The founder doesn't disappear in this version. They just stop being employed as human middleware. Judgement stays with the person who has money, reputation and possibly their house riding on the outcome, while the machines get on with the work they're actually good at.  Which leaves only one unresolved problem with the AI native company:

Finding the tab that's playing music.