Most engineering orgs are running cheap-capital playbooks they no longer have the cheap capital for

Viktor Cessan opens his essay with a sentence that should sit on every founder’s wall: Software development is one of the most capital-intensive activities a modern company undertakes, and it is also one of the least understood from a financial perspective. The gap, he argues, isn’t a personal failing of any one CFO or VP Eng. It’s structural — built over two decades when capital was cheap enough that the question is this team generating 3-5x its cost in returned value could be deferred indefinitely. Activity was a fine substitute for value while money was free.

Then capital became expensive again in 2022, and the substitution didn’t auto-correct. The behavior pattern was never connected to financial logic in the first place. Teams that measure success by features shipped — Cessan’s specific example — will always have something to show. Teams that measure success by return generated will sometimes have to report that it does not know. One of those framings is comfortable to defend in budget reviews; the other is honest. The comfort-honest gap defines a whole class of orgs running on momentum.

The AI moment Cessan flags is what makes deferral no longer viable. Accumulated complexity used to be an asset — the moat that kept competitors out. Then Nathan Cavaglione replicated 95% of Slack’s core functionality in fourteen days using AI agents. Work that took thousands of engineers over a decade. The moat became sand. That isn’t a story about AI replacing engineers; it’s a story about why your accumulated complexity now has to be financially justified going forward, because if your team can’t articulate why the complexity earns its keep, someone with a Claude subscription will articulate it for you by replicating 95% of your product over a weekend.

For a small team or indie founder, Cessan’s argument compresses to one diagnostic question: do you know what your last six months of engineering hours produced, measured in revenue or in clearly-attributable user retention? If the answer is a lot of features, you’re running the cheap-capital playbook on small-team scale. You can probably still get away with it for another year. But the calculation is on the table now in a way it wasn’t in 2020, and the founders who can answer the question will outcompete the founders who can’t — not because they’ll work harder, but because they’ll cut differently when they have to cut.