Fetching from the wire…
01
02
03
04
05
06
07
08
Public story · 2026-08-03 · high
One engineer skipped the rulebook for a two-week Swift rewrite, checking the agent's work with a screenshot diff instead.
Why now: Cherny's port and the HANDBOOK.md numbers surfaced in the same week's coverage, which is what makes the connection between them visible.
Cherny ported a full Electron app to Swift over two weeks, checking the rewrite with a screenshot diff against the running original.
The stakes: just 36.2% compliance with written policy alone, per the HANDBOOK.md paper on arxiv.
HANDBOOK.md's own failure analysis names the mechanism: details lost over extended sequences. A rule stated at the start of a run has to survive to the end, and the compliance number says it usually doesn't.
Cherny's fix skips the written rule altogether. Instead of an instruction to match the existing UI, the agent runs a screenshot comparison against the live Electron app itself. There's no prose to drift from because there's no prose in the loop.
The same swap generalizes past UI work: replace a binding prose rule with a hook or tool-level assertion that fails the turn when it's violated. An assertion doesn't degrade as a transcript grows. Prose does.
That's the bet worth watching: teams that swap prose rules for scripts should see their compliance numbers hold steady as runs get longer. Teams that don't should watch their numbers drift toward HANDBOOK.md's 36.2%.
Each link below shares sources, entities, or timing with this story.
Shared entities / Same source domain / Shared topic
Both cover Cherny, Electron, Replace, Swift; reported by the same outlet (arxiv.org); overlapping topics (against, agent, cherny).
Shared entities / Same source / Shared topic / Earlier coverage
Both cover HANDBOOK, Prose; cite the same source (HANDBOOK.md compliance results); overlapping topics (against, agent, compliance, policy).
Shared entities / Same source domain / Shared topic / Earlier coverage
Both cover Give, Prose; reported by the same outlet (arxiv.org); overlapping topics (against, agent, prose).
Shared entity: Give / Same source domain / Shared topic / Earlier coverage / Tension
Both cover Give; reported by the same outlet (arxiv.org); overlapping topics (agent, autonomou).
Swift uses Metal / Shared entity: Swift / Shared topic / Earlier coverage
Linked by a graph relationship (Swift uses Metal); both cover Swift; overlapping topics (against, binding).
Shared entity: Replace / Shared topic / Earlier coverage / Tension
Both cover Replace; overlapping topics (against, agent, autonomou); earlier Replace coverage from 2026-02-22.
Shared entity: Prose / Shared topic / Earlier coverage
Both cover Prose; overlapping topics (agent, check, compliance, prose); earlier Prose coverage from 2026-07-28.
Shared entity: Replace / Same source domain / Shared topic / Earlier coverage
Both cover Replace; reported by the same outlet (arxiv.org); overlapping topics (agent, binding).