The default path in startups is heroic: one founder, one idea, all in. Raise money, hire a team, spend two years building, and pray the market agrees with your thesis. Sometimes it works brilliantly. Most of the time it does not. The hit rate for venture-backed startups sits somewhere around 10%. For bootstrapped ones, it is worse.
We looked at those numbers and decided to build a different kind of machine.
The single-shot problem
Here is what nobody tells you about the "pick one idea and commit" model: it optimises for narrative, not for learning. The founder who raises $2M for a single product has enormous pressure to make that specific product work. Pivoting feels like failure. Killing it feels like career death. So they keep pushing, adding features nobody asked for, tweaking positioning, running one more experiment -- long after the data has spoken.
The sunk cost is not just financial. It is emotional, reputational, and temporal. Two years disappear. The founder emerges exhausted, possibly wiser, but starting from scratch.
Now imagine a different structure. One where killing a bad idea in week six is not failure -- it is the system working correctly. Where every dead end sharpens the next hypothesis. Where the infrastructure, the playbooks, the customer relationships, and the AI tooling carry forward to every subsequent build.
That is a venture studio. And that is what we are building.
Why now, specifically
Venture studios are not new. Idealab was doing this in the late 1990s. Rocket Internet scaled the model across emerging markets. What is different now is AI.
Before AI, a studio's advantage was mostly operational: shared back-office, reusable tech stacks, faster hiring. Useful, but incremental. The actual hard work of researching a market, building a prototype, testing with customers, and iterating on the product still required large teams and long timelines.
AI has changed the cost structure of that work by an order of magnitude. A two-person team with sharp judgment and good AI tooling can now do in a week what used to take a team of eight a quarter. Market research that consumed an analyst for a month gets synthesised in days. Prototypes that took sprints ship in hours. Customer interview analysis that required a dedicated PM happens in real time.
AI did not make studios interesting. It made them inevitable.
What we are actually building
Not an incubator. Not an accelerator. Not a holding company that slaps its logo on other people's ideas.
We build our own ventures from scratch. We start with a problem, validate obsessively, build fast, and either scale or kill based on evidence. Every venture runs through the same disciplined process -- research, MVP, iteration, scale -- with strict stage gates between each phase. If a venture does not earn the right to continue, it does not continue. No politics. No sunk-cost negotiations.
The studio itself is the product. Each venture makes the studio smarter: the research gets sharper, the prototyping gets faster, the pattern recognition compounds. A failed venture in insurance teaches us something that makes the next fintech build better. Not because the domains overlap, but because the process of finding product-market fit has transferable structure.
The leverage equation
Think about what a studio actually compounds:
Judgment. After validating dozens of hypotheses, you develop an instinct for which problems are real and which are founder delusions. You learn to read the gap between what customers say and what they do. This cannot be taught in a course. It comes from repetition.
Infrastructure. Every venture we build deposits code, tooling, workflows, and operational playbooks back into the studio. The second venture starts from a better baseline than the first. The tenth starts from a dramatically better one.
Speed. AI handles the throughput -- research synthesis, code generation, experiment scaffolding, data analysis. Humans handle the hard decisions. This division of labour means a small team can run at a pace that would have required a department five years ago.
Optionality. Instead of betting everything on one thesis, we can validate several in parallel and concentrate resources on the one with the strongest signal. This is not hedging. It is disciplined search.
What this is not
This is not a "throw things at the wall" operation. Velocity without direction is just expensive chaos. Every venture starts with a clear thesis: a specific problem, a specific customer, a structural reason why AI gives us an unfair advantage in solving it. We do not build things because they sound interesting. We build things because the evidence says someone will pay for them.
It is also not a lifestyle business. The goal is to build companies that scale independently -- with their own teams, their own customers, their own momentum. The studio is the launchpad, not the ceiling.
The honest version
We built a venture studio because we looked at the alternative -- pouring years into a single bet with a 90% chance of failure -- and concluded there had to be a smarter way to find what works.
Not smarter as in cleverer. Smarter as in structurally sound. A system that treats failure as data, compounds every lesson, and uses AI to collapse the time between hypothesis and evidence.
That is [r]think. A machine for finding problems worth solving -- and then solving them with unreasonable focus.