FIELD NOTE / TIKTOK
A reviewer can inherit the author's favorite proof.
The short film, the complete written thought, and the evidence behind it.
The TikTok edition will be linked here after its public post is verified.
A reviewer can inherit the author's favorite proof.
Day 07 · 2026-10-04 · TikTok
Short video caption
A second agent rerunning the author's favorite test is weak review. Give it the original promise, separate evidence and a protected hidden-state probe; require a deliberate mapping fault to fail. Simulated example. #EricFieldNotes
Full written post / accessible read
An agent builds a feature and shows a tidy passing test. A second agent reviews the patch and reruns exactly that test. It agrees. Did anything independent happen?
In this simulated case, both check a success page while the persisted account tier is wrong. Their conversation sounds rigorous because the test really passes. The hidden state was never part of either agent's exam.
The reviewer receives the original user promise and a protected read-only outcome probe. It is not shown the author's chosen test until after it writes its own failure hypothesis. A deliberate tier-mapping fault must make the hidden fixture fail.
Give review agents different evidence and require a negative control before counting agreement. Keep the actual verdict in a protected runner. Do this because rerunning one favored test is confirmation, not adversarial review.
#EricFieldNotes
Evidence and boundary
On-screen boundary: SIMULATED AGENT REVIEW. The sources below support documented mechanisms and specifications; illustrative scenarios are not presented as measured incidents.