← sources

Are coding agents creating a review wall, or am I inventing a problem?

source: reddit (reddit:indiehackers)published 2026-09-23 10:45ingested 2026-09-24 02:37

retained text

Over the past couple of weeks I’ve had multiple discussions with PMs and eng leads at companies of different sizes about how hard it’s been for engineering teams to transition to AI safely and effectively.

The fears are incidents, slop, review bombs that burn people out, missing guardrails, and how expensive it gets without a clear return.

I can see agents producing code (often better than most of us), tests, and PRs faster than any team could before, but the rest of the engineering process hasn’t caught up. Requirements are weak. Architecture is poorly documented and hard to navigate: it still lives in people’s heads instead of anywhere an agent can read. Quality still depends on someone reviewing every pull request. So the faster agents write, the more review becomes the wall everything backs up against. The worst part: more people are just rubber-stamping approvals, and incidents climb with it.

Median time to first PR review is up 156.6% (just read this: [https://hai.stanford.edu/ai-index/2025-ai-index-report/responsible-ai](https://hai.stanford.edu/ai-index/2025-ai-index-report/responsible-ai)). Teams spend nearly double the time in review overall. That’s insane and at the same time it makes sense…

I’m thinking about building something in this space (some sort of a Code Factory). Rough idea: a controlled loop with optional human checkpoints, agents for each phase, GitHub PR as the source of truth for the diff. Stop being the glue between Plan and Merge. Something that improves the codebase and measures progress, so little by little you can release humans from the babysitting.

Before I sink months into it, I’d rather know if the pain is real and if people would actually adopt.

If you’ve been there:

1. What’s the most annoying part of running coding agents on real product work right now?
2. What do you already use as a workaround (manual chats, kanban, GSD-style workflows, custom scripts, nothing)?
3. Would you switch if something handled Plan → Implement → Test → Review → Merge, you still owned merge, and you could confidently step away from each phase a bit at a time?
4. What other problems in this space am I missing?

Happy to hop on a short call if this is eating your week or DMs are open. And if this is a solved problem and I’m late: tell me what I’m missing.

Thanks a lot for any honest pushback.

document fetched 2026-09-24 00:29 · canonical https://reddit.com/r/indiehackers/comments/1wo2ilu/are_coding_agents_creating_a_review_wall_or_am_i