Aviator ads

They opened a PR. They requested review. That was 412 days ago. Some PRs never come back from the review queue. Don't let yours become skeletons.

Move the human review upstream. Engineers review intent and acceptance criteria before a single line is generated, not 500-line diffs after.

I have nothing to add 😁 Thanks Dan Robinson Aviator MergeQueue - > https://lnkd.in/gfur2rvW

Do you have an anti-AI-slop registry at your engineering org? It's a fundamental part of Aviator Verify: https://lnkd.in/gS96kreX These are the rules that apply to every change in your organization, and we call them org invariants. Every codebase has patterns that AI consistently gets wrong. Maybe it’s a specific error-handling convention, a naming pattern, a module boundary that gets violated, or a library the AI prefers that you’ve deprecated. These are knowable and preventable. Invariants are how you codify what these are for your specific codebase. These rules are defined once and checked automatically. Developers don’t need to include them in specs, they’re always enforced. As AI generates more of your code, the review burden shifts. You can't manually catch the same mistake 50 times across 50 AI-generated PRs. You need the pattern documented, automated, and enforced.

When we built Aviator Merge Queue (https://lnkd.in/gfur2rvW), agentic coding was not a thing. We built it because merging code at scale was already painful for large engineering teams. But something interesting is happening now: teams adopting agentic coding are reaching out to us specifically because they're running into merge conflicts they can't keep up with. When multiple agents are working in parallel on the same codebase, merge conflicts stop being a nuisance and start being a delivery bottleneck. Stacked PRs and merge queue turn out to be a pretty natural fit for this problem. Stacked PRs break large agent-generated changes into smaller, reviewable chunks. Merge queue serializes those changes intelligently so parallel workstreams don't constantly collide on main. Neither of these was designed with AI agents in mind, but the underlying problem they solve — coordinating parallel changes to a shared codebase — is exactly what teams are struggling with now. What I find interesting is that agentic coding is reinforcing some very old software engineering fundamentals. Clear ownership, small incremental changes, disciplined merging.

Move the checkpoint upstream. Engineers review intent and acceptance criteria before a single line is generated, not 500-line diffs after.