Constraint Was Doing the Deciding

Why the second film usually disappoints, and what that has to do with an engineering team that just got a blank check.

Something I noticed in film school took me until this year, and an AI-assisted engineering team, to fully understand.

A director's first feature is often made on almost nothing. Every scene has to fight for its place, because every scene costs money nobody has. Lines get cut because a location is unaffordable. Whole subplots die in a production meeting because there is one day left to shoot. And when those films work, they tend to work astonishingly well: lean, specific, and strangely inevitable, as if nothing in them could have been otherwise.

Then the director gets hot. The next film arrives with a real budget, a bigger cast, and a studio that says yes. And the second film, more often than anyone likes to admit, disappoints. Longer, looser, full of things that are individually fine and collectively shapeless.

The easy explanation is that the first one was a fluke. I don't think that's it. I think the budget was doing the editing.

The shark that didn't work

The most famous version of this is Jaws. The mechanical shark kept failing, so Spielberg shot around it, suggesting the shark with yellow barrels, point-of-view shots, and a two-note score instead of showing it. You barely see the thing for much of the film, and the film is far more frightening for it. A malfunction made a decision nobody would have made voluntarily, and it was the right one.

Jaws was, technically, Spielberg's second theatrical feature, which is the whole point. The number on the film was never what mattered. The budget was.

That is the part that stuck with me. Constraint limits what you can make, but its real work is forcing decisions. Under a hard budget, every trade-off has to be argued, and the arguing is where the quality comes from. Take the budget away and the decisions still need making. What disappears is the pressure that got them made.

We just got the blank check

For my entire career, engineering capacity was the budget. Every feature had to fight for a slot in the sprint. Every idea had to justify the weeks it would consume. Plenty of good ideas lost, and I resented that, but the fight meant that whatever shipped had survived an argument.

This year that budget mostly disappeared. My team ships more than two hundred tickets a month. A first version of an entire application can now come together in days. By any historical standard, we are the hot director with the studio saying yes.

I expected that to make our output better. What actually happened was stranger.

My ideas didn't improve. There are just more of them. The ratio held: the same share of good ideas, the same share of merely fine ones, the same share that should never have been built. Throughput went up and the mix stayed where it was.

That is worse than it sounds, because the dangerous category was never the bad ideas. Bad ideas die on their own; someone notices. The dangerous ones are the plausible ones, the ideas that are fine. Under a hard budget, fine loses to good in the fight for the slot. Without the budget there is no fight, and fine rides straight into the cut alongside everything good. That is what second films are made of.

Why you never have enough engineers

Every engineering leader already knows a version of this, even if they have never framed it this way. You never have enough engineers. Double the team and the roadmap doubles to meet it. Within a quarter you are constrained again.

That used to feel like a staffing problem. I now read it as evidence that the constraint is structural. It never goes away. Remove it in one place and it reforms somewhere else.

AI removed the build constraint, and the constraint moved downstream, into the places nobody reviews at planning time: support load, the customer's capacity to absorb change, maintenance, and the attention of the few people who make judgment calls about all of it. The constraint still arrives. It just arrives after the merely fine ideas have already shipped, which is the most expensive possible moment to meet it.

Manufacture the budget

So the move is to put the constraint back on purpose, at the point where it can still make decisions, instead of waiting for it to find you downstream.

I wrote recently about rationing decisions, a decision budget, to protect the quality of the ones you make. This is why that works. A self-imposed budget is a manufactured hard budget. It restores the forcing function that old constraints used to supply for free.

Once you look for these, they show up in small places. Our development workflow won't advance anything the person responsible can't explain, which forces the same argument a hard budget used to force. When I made logos for our internal tools recently, I held every one of them to a shared palette and a single mark, and that constraint is why they read as a family instead of five unrelated pictures, and why other leaders can now make their own without fragmenting the brand. And this essay exists because of one. I have far more ideas than weeks, and one slot a week is the only thing that makes me choose which of them deserves writing.

I impose every one of those constraints myself, because nothing else is going to anymore.

The first film is great because every frame has to justify itself. We are all making the second film now, with more capacity than any of us has ever had. The question worth asking is whether you will write your own budget, or let the blank check make the film for you.