feat(growth): add inspectable company research pilot - #1009
Merged
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
Contributor
Contributor
blove
enabled auto-merge (squash)
September 5, 2026 16:57
Contributor
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Adds a local, company-only research pilot that compares a bounded Dawn agent with the existing enrichment generator on identical captured evidence. Operators can inspect immutable run records, rejected findings, source capture failures, usage, and blinded human-review packets. The managed synthetic deployment remains isolated, and no Growth contacts or email delivery are connected.
Adds a scoped Node 24 CI lane and required-check gate for Growth Research; previously its project tag selected no verification job. The 130 CI classifier/workflow contract tests pass, including rejection of an in-scope skipped job.
The agent uses Dawn’s supported authored Zod schema to represent unknown fields correctly on 0.8.24. Dogfooding also produced the upstream nullable-schema fix in cacheplane/b4run#573.
Validation: growth-research tests, check/typecheck, lint and build passed; independent spec and code review passed. Live agent runs structurally completed 6/6 synthetic cases and 12/12 public-company repetitions, including empty-evidence abstention. Anthropic billing rejected the baseline calls, so comparative quality and human scoring remain pending; no claim of improved factual accuracy.