Hacker News front pageSep 26, 2026, 23:06 UTC+8

Code review is more than detection: John Allspaw pushes back on agent-driven replacement

There is more to code review than (automatable) detection

John Allspaw of Adaptive Capacity Labs responds to a paper arguing coding agents can replace human code review. He says the paper reduces review to four automatable functions and misses what humans actually bring: genuine confusion as a signal, questioning whether a change is needed at all, noticing what is missing, calibrated scrutiny based on who wrote the code, and coactive knowledge-building. He also points out that reviewers carry operational context and personal accountability that agents lack. The post is a point-by-point rebuttal of the paper's framing; no quantitative experiments are presented.

Why it matters: John Allspaw delivers a systematic rebuttal to a paper claiming AI can replace human code review, listing seven layers of cognitive work that resist automation — arguments are concrete and grounded in operational experience. Hits all three HKR axes, but as an opinion piece rat...

Read the original ↗Export Markdown

Take

This is worth reading because the rebuttal hits where it hurts. The paper decomposes code review into four functions—defect detection, style enforcement, knowledge transfer, awareness—and argues agents can do each one, therefore agents can replace humans. Allspaw doesn't argue about whether agents can do those things. He says the decomposition itself is wrong: the stuff that matters isn't on your list.

Anyone who's done real engineering will recognize his examples. A senior engineer saying "I don't understand this" is itself a finding—the code is too convoluted or the abstraction is wrong. An LLM will always "process" the code; it can't give you that signal of genuine human incomprehension. Questioning whether a change is even necessary, noticing that an API contract changed but error handling didn't, calibrating scrutiny based on who wrote the code—none of this fits into "detection."

This is a pure opinion piece, no experiments. But Allspaw is former CTO of Etsy with deep background in resilience engineering and incident analysis, so his pushback isn't armchair commentary. If you're thinking about automating code review, read this—not because it gives answers, but because it draws the line where automation stops making sense.

Source: Hacker News front page · adaptivecapacitylabs.com
Published: Sep 26, 2026, 23:06 UTC+8 · 3 days ago