Andrew Dinh

4% of GitHub public commits are being authored by Claude Code right now. At the current trajectory, we believe that Claude Code will be 20%+ of all daily commits by the end of 2026. While you blinked, AI consumed all of software development.

62

4 Comments

GitHub's framing: we are in the "Eternal September" of open source. AI-generated contributions are overwhelming OSS maintainers. Volume is up dramatically. Review costs have not dropped. The result: maintainer burnout, slower reviews, and degraded quality signals for the projects your production stack depends on. 𝗪𝗵𝗮𝘁 𝘁𝗵𝗶𝘀 𝗺𝗲𝗮𝗻𝘀 𝗳𝗼𝗿 𝗖𝗧𝗢𝘀 𝘄𝗵𝗼 𝗰𝗼𝗻𝘁𝗿𝗶𝗯𝘂𝘁𝗲 𝘁𝗼 𝗼𝗿 𝗱𝗲𝗽𝗲𝗻𝗱 𝗼𝗻 𝗢𝗦𝗦: 𝗧𝗵𝗲 𝘂𝗽𝘀𝘁𝗿𝗲𝗮𝗺 𝗿𝗶𝘀𝗸: Projects in your dependency tree—including foundational ones like Node.js, Python, curl, Apache—are receiving a surge of AI-generated PRs. Maintainers are spending more time triaging noise. The projects that keep your infrastructure running are under strain. 𝗬𝗼𝘂𝗿 𝘁𝗲𝗮𝗺'𝘀 𝗔𝗜 𝗰𝗼𝗻𝘁𝗿𝗶𝗯𝘂𝘁𝗶𝗼𝗻 𝗯𝗲𝗵𝗮𝘃𝗶𝗼𝗿 𝗺𝗮𝘁𝘁𝗲𝗿𝘀: If your developers use AI coding tools to generate open source contributions, the quality of those contributions affects your relationship with maintainers and your team's reputation in communities you depend on. GitHub is shipping tooling in response: - Contributor signals: reputation scores to help maintainers prioritize review - Agentic triage workflows: AI-assisted triage for maintainers to process volume - Configurable contribution gates: maintainers can set requirements for new contributors 𝗧𝗵𝗿𝗲𝗲 𝗽𝗼𝗹𝗶𝗰𝘆 𝗮𝗰𝘁𝗶𝗼𝗻𝘀 𝗳𝗼𝗿 𝗲𝗻𝘁𝗲𝗿𝗽𝗿𝗶𝘀𝗲 𝘁𝗲𝗮𝗺𝘀: 1. 𝗘𝘀𝘁𝗮𝗯𝗹𝗶𝘀𝗵 𝗲𝘅𝗽𝗹𝗶𝗰𝗶𝘁 𝗔𝗜 𝗰𝗼𝗻𝘁𝗿𝗶𝗯𝘂𝘁𝗶𝗼𝗻 𝘀𝘁𝗮𝗻𝗱𝗮𝗿𝗱𝘀. AI-generated code that ships to production via OSS PRs needs the same review bar as internal code. "The AI wrote it" is not a quality guarantee. 2. 𝗥𝗲𝘃𝗶𝗲𝘄 𝗰𝗼𝗻𝘁𝗿𝗶𝗯𝘂𝘁𝗶𝗼𝗻 𝗳𝗼𝗼𝘁𝗽𝗿𝗶𝗻𝘁. Know which OSS projects your team is contributing to and with what volume. Respect maintainer bandwidth as you would a supplier relationship. 3. 𝗖𝗼𝗻𝘀𝗶𝗱𝗲𝗿 𝘀𝗽𝗼𝗻𝘀𝗼𝗿𝗶𝗻𝗴 𝗱𝗲𝗽𝗲𝗻𝗱𝗲𝗻𝗰𝗶𝗲𝘀 𝘆𝗼𝘂 𝗿𝗲𝗹𝘆 𝗼𝗻. If your business depends on a project, the maintainer's time has economic value to you. GitHub Sponsors and direct foundation support are direct ways to invest in that supply chain. The OSS commons is infrastructure. Its degradation is a supply chain risk. Does your team have an explicit AI contribution policy for open source projects? What standards do you hold AI-generated OSS contributions to? #OpenSource #CTO #SupplyChain

Fatbobman Xu examines why developers stick with Core Data, citing mature capabilities, existing projects, and its object graph design. He notes growing mismatches with modern Swift in expressiveness, concurrency, and type safety, and argues experience-based workarounds are unsustainable. Read the article at https://lnkd.in/dAcDwfkV

1

If you're using agents such as Claude Code, Codex, or Cursor, there's a powerful new skill that will help you write smarter, simpler, and more modern SwiftUI 🚀 https://lnkd.in/g3swmDmd

5

Most agent benchmarks ask: “Does the code compile?” We evaluated a stricter question: Did the agent use its resources correctly? Across 300+ real npm and PyPI libraries, giving agents structured, versioned library context (tiles) led to: 📈 ~1.2X improvement overall in correct, idiomatic API usage 📈 Up to 3.3X improvement for recently released libraries We tested for common agent failure modes, using day-to-day eval scenarios. Check out the evals and results in the Tessl Registry. 🔗 Link in comments.

15

1 Comment

FYI: Astral has just launched a standalone repository for their new type checker, Ty, on GitHub. Ty is an extremely fast Python type checker written in Rust. https://lnkd.in/g55FMUru #Python #TypeChecking #Rust #Programming #DeveloperTools

8

- August Approved Grants Updates - S5 v1: Rewrite in Rust added guides, CLI tools, and new crates while refactoring store types for better ergonomics. Also introduced encrypted p2p networking via Iroh. Upcoming work includes finishing the RHP4 proxy, implementing missing specs, and adding end-to-end encryption. https://lnkd.in/e6tWZxt9

1