rfc: dedicated review time and stacked PRs - #18
Open
hannahhoward wants to merge 1 commit into
Open
Conversation
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
bajtos
approved these changes
Aug 3, 2026
|
|
||
| We are currently moving faster than we ever have as a development team. Since June 1, we've merged 221 PRs. The median PR review process is healthy: 6 files, merged inside two days. It's the tail that's the problem: 34 merged PRs touched 30+ files, 12 touched 60+, and p90 time-to-merge is 11 days. Unfortunately, this tail covers our most important work: ucantone #30 (attested signatures, 114 files), ingot #35 and #36 (Hilt integration, 103 files each), piri #22 (Curio PDP pipeline, 86 files). | ||
|
|
||
| What's wrong? In short, building is more fun than reviewing, AI has made building a lot more fun and a lot more fast. I am as guilty as anyone. So our bottleneck is often review and corresponding context management -- having to keep up with the rapid pace of new code committed and new ideas explored. |
Member
There was a problem hiding this comment.
I am as guilty as anyone.
So am I!
Having written that, I feel this sentence does not belong in an RFC. Let's remove it.
Suggested change
| What's wrong? In short, building is more fun than reviewing, AI has made building a lot more fun and a lot more fast. I am as guilty as anyone. So our bottleneck is often review and corresponding context management -- having to keep up with the rapid pace of new code committed and new ideas explored. | |
| What's wrong? In short, building is more fun than reviewing, AI has made building a lot more fun and a lot more fast. So our bottleneck is often review and corresponding context management -- having to keep up with the rapid pace of new code committed and new ideas explored. |
|
|
||
| ## Proposal 1: dedicated review time | ||
|
|
||
| Each of us holds one hour a day on the calendar for review, checks the review queue before starting new build work, and gives a first response to any review request within one business day. Google's measured cost is about 3 hours a week, so an hour a day is deliberately above steady state: we have a review backlog to drain and agents multiplying our output. Revisit the number in a month. |
Member
There was a problem hiding this comment.
I agree with everything proposed here.
I suggest these additions to consider:
- Configure CODEOWNERS in each repository, so that every new PR automatically requests review from at least one person.
- Monitor "time to first response to a review request" on a per-user and team-wide basis. Every week, an automation should post a message to our eng Slack channel with a list of p50 & p90 for each team member.
- Implement basic alerting. For example: at noon in my local time, if there are any outstanding review requests from yesterday or before, send me a Slack DM with the list of review requests needing my attention.
Member
There was a problem hiding this comment.
I agree with Miro, simple slack reminder would be great.
Comment on lines
+34
to
+36
| A PR should be reviewable in under an hour, which the evidence puts at roughly 400 changed lines, though the content is just as important -- critical auth logic takes longer than code-gen'd serialization for example. Therefore, bigger work must ship as a stack of dependent PRs, each one a reviewable step. | ||
|
|
||
| The norm binds our agents too. Each active repo's AGENTS.md gets an instruction that work expected to exceed the guide ships as a stack, and Claude Code gets the stack tooling below. |
Member
There was a problem hiding this comment.
Is it worth adding some automation here as well? E.g. a GHA workflow that posts a warning comment to the PR discussion if the diff is larger than 400 lines.
pyropy
approved these changes
Aug 3, 2026
pyropy
left a comment
pyropy
left a comment
Member
There was a problem hiding this comment.
Great initiative, I agree with everything in this RFC.
alanshaw
approved these changes
Aug 3, 2026
|
|
||
| ## Decisions needed | ||
|
|
||
| - **The hour.** I propose one hour a day. Google's measured average is about three hours a week, so this is deliberately generous while the backlog drains. |
Member
There was a problem hiding this comment.
Agree as a baseline, I spend significantly more than this daily already though.
| ## Decisions needed | ||
|
|
||
| - **The hour.** I propose one hour a day. Google's measured average is about three hours a week, so this is deliberately generous while the backlog drains. | ||
| - **Pilot now vs. wait.** The GitHub preview is days old and will have sharp edges; spr covers the gaps. I propose pilot now. |
|
|
||
| - **The hour.** I propose one hour a day. Google's measured average is about three hours a week, so this is deliberately generous while the backlog drains. | ||
| - **Pilot now vs. wait.** The GitHub preview is days old and will have sharp edges; spr covers the gaps. I propose pilot now. | ||
| - **Size guardrails.** Social norm only, or a soft CI warning above ~400 changed lines? I lean social to start. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
📖 Preview
Summary
Two commitments to fix our review bottleneck: an hour a day of calendar-held review time with a one-business-day first response, and stacked PRs for anything bigger than an hour's review — recommending GitHub's native stacked pull requests (with our Claude Code workflows taught to produce stacks), spr as fallback, Graphite in reserve.