Don't get too far ahead of the agent

When I started Sherwood I designed the whole back-end up front, before I’d built any of the front-end. It felt productive. When it came time for the front-end I expected that the agent would just be able to code until it was done, but by the time we got there, the back-end had drift in it that had to be reconciled. You hear a lot of the term ‘Human in the Loop’ posed as a solution for this problem, and it’s true that it helps, but this also slows development, and I don’t want to be a bottleneck. Right now the issue is that agents can write much faster than one person can review, so keeping pace will eventually require some kind of automated review where humans only have to intervene when needed.

Post-note: the industry is starting to call a version of this spec-driven development; write the spec, let the agent build to it. Whether today’s agents can actually keep drift in check without that human in the loop is still an open question.