Writing on venture building, product and the studio model.

There’s a version of venture building that looks very tidy from the outside. Clear stages. Neat diagrams. Validation gates you pass through like airport security.

I’ve spent the last six months inside a real build, and I can tell you: that version is mostly fiction. Not malicious fiction. Comforting fiction.

Because if venture building were just a matter of following the right steps, we’d all sleep better. But the truth is less reassuring and far more interesting.

Where playbooks actually break

Playbooks work best when the problem is stable, the customer is well understood, and the cost of being wrong is low. Early-stage ventures tick none of those boxes. What actually happens is more like this:

• you inherit a set of assumptions that feel “validated” but wobble under scrutiny
• early user signals conflict rather than converge
• confidence in the solution starts to outrun confidence in the problem
• progress creates pressure to keep going, even when something feels off

This is the moment everyone quietly hopes the framework has an answer. It doesn’t. Because no playbook can tell you whether this assumption matters more than that one, or whether slowing down now will save you three months later. That’s a judgement call. And it’s on you.

Judgement is not instinct (and it’s not vibes either)

This is important: judgement isn’t gut feel. Good judgement is built from pattern recognition, what you’ve seen before; signal detection, what’s actually new here; and constraint awareness, what you can afford to get wrong. It’s the ability to ask:

• What would have to be true for this to work?
• What are we pretending we know?
• If we’re wrong, how will we find out fastest?

And then make a call anyway. This is why two teams can follow the same process and get wildly different outcomes. The process is shared. The judgement isn’t.

Process answers “what next?” Judgement answers “should we?”

Most startup tooling is built around forward motion: next sprint, next build, next test. Judgement lives one level above that. Judgement is what stops you from:

• building the wrong thing very efficiently
• validating the wrong question extremely well
• mistaking activity for progress

Some of the hardest calls I’ve seen (and made) weren’t about what to do next. They were about whether we were asking the right question at all. Those decisions don’t feel productive in the moment. They often look like slowing down. They rarely get applause. They also tend to be the ones that matter most.

Autonomy without shared beliefs is a heavy lift

Studios love the idea of autonomy. I do too. But autonomy without shared beliefs about how decisions should be made creates a strange kind of pressure:

• you’re empowered, but unsure of the edges
• accountable, but unclear on what “good” looks like
• encouraged to move fast, but corrected after the fact

That’s not a character flaw. It’s a system gap. What helps isn’t more oversight. It’s shared decision principles: what are we optimising for at this stage, what level of validation is “enough for now”, and when is speed irresponsible?

Without those, builders don’t just make decisions. They also carry the anxiety of whether those decisions will later be judged against a different, unstated standard. That’s a lot of load for one role to absorb.

A practical test for better judgement

Momentum feels good. Reduced uncertainty compounds. Good judgement usually leaves you with a sharper problem statement, clearer trade-offs, and stronger conviction, even if the answer is “not yet”. Bad judgement often leaves you busy, slightly uneasy, and telling yourself you’ll “validate it properly later”. Later is expensive.

The work you don’t see is the work that counts

A lot of judgement work happens off the roadmap. It’s choosing not to pass every wobble onto the founder. Holding uncertainty so someone else can stay focused. Pushing back gently when confidence is running ahead of evidence. Making space for thinking when delivery is louder.

None of this ships. All of it shapes what eventually does. If venture building were purely mechanical, this stuff wouldn’t matter. But because it’s human, it does.

So what is the job, really?

The job isn’t to follow the playbook perfectly. It’s to make decisions with incomplete information, stay honest about what you don’t know, protect the fundamentals when pressure is highest, and adjust course without blowing up confidence.

And if that sounds abstract, try making these calls with real people, real money, and real consequences, while everyone’s watching. You’ll realise pretty quickly: judgement isn’t a soft skill. It’s the whole job.

A thought experiment, not a product. Nothing here was built. I’m writing it up because the question underneath it is the one I worked on at BfB Labs: what actually moves a habit.

Restriction makes opponents

Parental controls are built as restriction. Limits, blocks, timers. That makes the parent and the child opponents, and it leaves the behaviour itself untouched. The child learns to work around the rule, not to change the habit.

So the question was simple. Could the same data drive a behaviour change loop instead?

The loop

It’s built on the habit loop: cue, action, reward. A parental layer turns real behaviour into plain signals, and those signals become the cues. A game layer turns each one into an immediate reward, in a town the child builds and owns.

Onboarding is a pact the parent and child sign together. The agreement is the first cue, and it’s made side by side rather than handed down.

The hard part was failure

Every loop needs a failure state, and the obvious one is wrong. Taking away what a child has built is punishment, and punishment breaks the loop. The child disengages, and the thing driving the behaviour is gone.

Instead, a storm settles over the town, and only positive actions clear it.

The same loop as Oshn

It’s the same mechanism as Oshn, the product I led at BfB Labs, from the other end. Get the action to happen before the motivation arrives, then let the reward reinforce it. Oshn was a regulated product tested with clinicians. This was a thought experiment. The loop underneath is the same.