Fetching from the wire…
Public story · 2026-09-02 · high
Vercel's AI SDK says its own agent authors up to 35% of merged pull requests, and GitHub added Copilot approvals to the merge gate the same day.
Why now: Latent Space's maintainer interviews and GitHub's Copilot approval feature both posted on September 1, 2026.
Flue and tldraw now auto-close pull requests from outside contributors, converting them into issues instead of reviewing the code. The change matters most for first-time contributors, who now compete with a maintainer's own tuned agent for the same merge decision.
Vercel's AI SDK team describes a similar split without the auto-close. Its agent factory authors 25 to 35% of merged pull requests and closes 70 to 80% of issues, according to Latent Space. The team put its backlog at more than 1,000 open issues and nearly 800 open PRs. That backlog is why they automated in the first place.
For a contributor, the fix stops being the valuable part of a submission. A clean reproduction with a failing test now beats a patch, because generating the patch is what the agent is for. Write the issue like a task spec, since the issue is what the agent works from.
The AI SDK numbers offer a template for maintainers considering the same move. About a third of merged code comes from an agent, with a human still holding the merge button. Getting there took tuning the agent's configuration, not flipping a switch.
GitHub posted the change in its changelog on September 1, the day Latent Space's piece went up. Copilot code review can now submit approvals that count toward a repository's required-approvals rule. The feature is off by default. It's configurable down to specific file paths, and it gets dismissed on new commits like a human reviewer's approval would.
Each link below shares sources, entities, or timing with this story.
Open issues on the AI SDK went from 1,022 to 844 in four weeks. That's the number I'd lead with if I were writing this post, and Vercel did. Vercel published the internals of the system it built to maintain the AI SDK, a repo with over 20 million npm downloads a week, 26,000+...
Ryan Lopopolo from OpenAI Frontier went on the Latent Space podcast and described something I've been circling around for months. His team of three engineers built Symphony, OpenAI's internal orchestration layer, as a million-line Elixir codebase. Not one line was written by a...
Three things happened this month that only make sense together. Agent Plugins 1.0 shipped co-signed by six competitors: AWS, Anysphere, Microsoft, OpenAI, Vercel and Google (GitHub Changelog). It makes skills-plus-MCP bundles portable across clients. OpenAI's August 11 Codex c...
swyx's synthesis of the AI Engineer World's Fair 2026 is the clearest framing I've read of where this all goes. The discipline moved from building agents to engineering the harness around them. Lilian Weng (now at Thinking Machines Lab) reframed her whole practice as "harness...
Anthropic released Claude Fable 5.1 on September 1. Claude Code v2.1.257 made it the default Fable model at 17:53 UTC that day, with a 1M-token context window, $10 per million input tokens, $50 per million output, and $0.25 per million on cache reads (claude-code CHANGELOG). B...
On July 1, the Godot Foundation rewrote its contributor guidelines to ban nearly all generative-AI code submissions and autonomous-agent PRs. New contributors with three or fewer merged PRs now need maintainer permission before submitting features or refactors. (The Register)...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.