The Great Flattening

There's a shift happening in how software gets built, and it's not really about the models getting smarter. It's about the backlog. For years, every product team carried a graveyard of good ideas that never got built because there weren't enough engineers or enough hours in the sprint. That excuse is evaporating.

When intelligence gets cheap enough, the backlog stops being a capacity problem and turns into a decision problem. A founder can spin up five versions of a feature overnight, test them against real usage, and throw away four without losing a quarter of engineering time. The constraint isn't headcount anymore. It's judgment about what's worth building at all.

The companies leaning into this look different on paper. Fewer engineers, sure, but a much bigger line item for token spend, since AI-assisted coding runs on inference the same way a factory runs on electricity. Shipping cadence speeds up because the loop from customer request to merged code gets shorter, sometimes down to hours instead of weeks. There's less translation happening between what a customer wants and what actually gets built, because the people writing the code are closer to the people asking for it, and the AI is doing the grunt work in between.

Call it a flattening. Layers of process that used to exist because writing code was slow and expensive are getting compressed, and the org chart that made sense for that world starts to look oversized for this one. Not every company will restructure around it, and plenty will keep hiring the way they always have. But the ones that don't adapt are going to look increasingly strange next to competitors shipping ten times as many experiments for the same budget.