Teams adopting coding agents report the same surprise: writing code stopped being the slow part, and nobody had planned for what came next.
Review does not scale the way writing does
A human reviewer's throughput did not change because an agent got faster. If anything it got worse, because reviewing code you did not write, and that no human wrote, takes more care, not less.
- Smaller diffs, always
- Agents that explain their reasoning, not just their output
- Automated review as a first pass, humans as the second
This is why the code review category has been the fastest-growing part of the directory this year.