# PR reviews are taking longer than ever, are you actually catching agent-written bugs or just going through the motions? — retained record

> Retained reddit record from reddit:indiehackers.

Canonical: https://highsignal.app/data/records/daca535fd52b1d83

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:indiehackers)  published 2026-09-17 14:04 ingested 2026-09-18 02:39

[original source ↗](https://www.reddit.com/r/indiehackers/comments/1wiuxio/pr_reviews_are_taking_longer_than_ever_are_you/)

## retained text



```
<disclaimer maybe some ppl find long post but just trying to understand pain is real or not …>

 I have been talking a lot lately from devs and eng managers: review queues are backed up, the diffs are huge, and nobody is confident the code is actually safe to ship.

It makes sense. Agent-generated code moves fast. But the review process didn't evolve with it. You're still squinting at diffs, still relying on someone to mentally simulate what the agent did, and still hoping the happy path covers enough.

A few questions I'm genuinely curious about:

1/How are you actually verifying agent-written code before it merges? Is it still just human review, or have you layered something else in?

2/Are you running into edge cases that passed review and broke in prod anyway?

3/And the bigger one: do you trust that your current testing setup catches behavior issues, not just syntax? Compliance gaps, silent failures on edge cases, non-deterministic outputs at runtime?

We're building toward deterministic verification at FetchSandbox specifically because of this pattern. But I want to understand how teams are handling it today before we get too deep into assumptions.

what's actually working for you, or what's broken. Both are useful..looking for some feedback ..
```



document fetched 2026-09-18 00:28 · canonical https://reddit.com/r/indiehackers/comments/1wiuxio/pr_reviews_are_taking_longer_than_ever_are_you
