GitHub

How can you tell whether an AI code reviewer is finding the issues that matter, and where it falls short? We built ReviewBench, a new open benchmark for evaluating code review agents. Its 219 pull requests across 19 languages reflect patterns from an analysis of 103.9 million GitHub pull requests. Findings are assessed against a shared rubric. Senior engineers independently reviewed the golden set, and their judgments agreed with ReviewBench 96.6% of the time. Explore ReviewBench, compare agents, and evaluate your own. https://lnkd.in/e8v9BU_d

Git activity on GitHub more than doubled between September 2025 and August 2026, reaching 473.3 billion events per month. In September, developers and agents made 7.38 billion commits, more than 5× the number a year earlier. In the first post of our deep-dive series, we explain why agent-scale development calls for a new Git architecture and how we’re redesigning the infrastructure to scale reads and writes independently. Internal benchmarks show up to 35× higher write throughput, while the principles behind the work keep reliability, governance, and familiar developer workflows at the center. https://lnkd.in/em-Tyvtx

2022 ➡️ 2026 Here's your sneak peek at this year's #GitHubUniverse programmable badge. Plus a look at how it has changed over the years. Grab your in-person pass and secure a badge ⬇️ https://lnkd.in/eJ-6W6SK

Join us for Open Source Friday with Madelyn Olson, Valkey co-creator, maintainer, and Technical Steering Committee member. Valkey is an open source, high-performance key-value data store used for caching, real-time applications, and workloads that need fast access to data. We’ll explore what it does, the story behind the project, and how its community is shaping its future. Learn more and contribute at https://lnkd.in/derxpAmT.

Getting help at work shouldn’t feel like another job. AT&T data scientist Natalie Gilbert is building an HR chatbot for the moments when people need support without another complicated process. We sat down with Natalie to hear the personal story behind her work. She shared how GitHub Copilot helps her review code and catch errors, giving her more time to understand the people she’s building for. Meet Natalie and see the work behind Ask AT&T for HR: https://msft.it/6003aWxgc

AI-focused software engineers earned up to 18.7% more than their non-AI peers at the same career level in a 2025 Levels.fyi analysis 📈 Explore three practical skills for building software that uses AI models 👇

What if one local AI model could understand your screen, transcribe meetings, and answer questions about what you’ve seen? Join us for Open Source Friday with Ayush Shekhar, creator of ScreenMind—an open source project that brings vision, audio, and reasoning together using a single Gemma 4 model, running locally on a GPU with as little as 4GB of VRAM. Explore ScreenMind and contribute on GitHub: https://lnkd.in/emVxCKRx

Git 2.56 is here. ⚡ The new git add --resolved stages only paths that are still marked as conflicted, and refuses to stage them if conflict markers remain. One merge-base case dropped from 0.68s to 0.01s. The highlights 👇 https://lnkd.in/dpx2F76v

GitHub.com has fully moved off CSS-in-JS to CSS Modules, resulting in 55% less server-side render time and 25% faster component init. https://lnkd.in/eWWhYkjz

When an agent makes a change to your code, you want to review it, run it, and see what changed. Great news: now you can do all three without having to leave the GitHub Copilot app. Learn about the built-in panels that can change the way you work with AI 👇 https://lnkd.in/e-i68aN3