Gates Get Thicker
The branch picture barely changes as a team grows. What changes is what stands in front of the merge — and once AI writes the code, output multiplies while reading speed stays put, so a queue forms in front of approval. Adding approvers does not catch up; narrowing where human attention is spent does.
01Gates Get Thicker
Concept at a GlanceWith several people there are several branches. Each forks off briefly and comes back to the trunk, and that shape does not change as the team grows.
What changes is what stands in front of the merge. At small scale, one person's approval holds that spot.
Once AI starts writing, a queue forms in front of it. Output multiplied several times over while reading speed stayed exactly where it was.
Adding approvers does not catch you up. Human review is not a resource you can scale, so narrowing where it is spent is the only thing that works.
So AI review goes in front. It filters mechanical defects and duplication first, acting as a funnel that reduces what ever reaches a person.
Not even everything that gets through needs human eyes. Only paths that hurt when they break — auth, payments, migrations — get read closely; the rest go to AI and CI.
Anything arriving at once gets lined up and verified one at a time. Each passed on its own, but nobody has checked the combined state.
At enterprise scale one new question appears: who wrote this code? The answer has to be recorded on the commit for an audit to have anything to read.
That marking earns its keep when undoing things too, since it lets you pull out one agent's work in a single pass.
Passing conditions differ per path as well. Where merging may be autonomous, where a person is required, and where AI must not touch at all — all pinned down as policy.
02 Understand It Simply
For EveryoneScale does not change the branch picture; it adds gates in front of the merge. Human review cannot be scaled up, so instead of adding more of it, AI review and CI narrow what a person has to read.
At small scale one person's approval guards the merge.
Once AI starts writing, that spot becomes the bottleneck, and since human review is not a resource you can scale, you narrow where it is spent rather than adding more of it.
Put AI review in front to filter mechanical defects, have people read closely only on paths that hurt when they break — auth, payments, migrations — and send the rest to AI and CI.
Line up whatever arrives at once in a merge queue so the combined state is verified one at a time.
Enterprise adds provenance on top: recording which commits an AI produced is what lets an audit be answered, and it is also what lets you pull one agent's work back out in a single pass.
- –Responding when PR volume outruns review capacity
- –setting per-path risk tiers and CODEOWNERS
- –judging when a merge queue is due
- –and recording the origin of AI-written commits
03 Frequently Asked Questions
FAQWhat is Gates Get Thicker?+
The branch picture barely changes as a team grows. What changes is what stands in front of the merge — and once AI writes the code, output multiplies while reading speed stays put, so a queue forms in front of approval. Adding approvers does not catch up; narrowing where human attention is spent does.
Where is Gates Get Thicker used?+
Responding when PR volume outruns review capacity, setting per-path risk tiers and CODEOWNERS, judging when a merge queue is due, and recording the origin of AI-written commits.
What's a simple analogy for Gates Get Thicker?+
Scale does not change the branch picture; it adds gates in front of the merge. Human review cannot be scaled up, so instead of adding more of it, AI review and CI narrow what a person has to read.