GitHub

Is AI making us all use the same tools, or is it empowering us to try new things? 🤔 The Head of GitHub Next, Idan Gazit, sees two trends colliding: • Consolidation around popular frameworks where AI excels • Lower barriers to programming languages you've never written What do you predict will win out? Gather more insights here. ⬇️ https://lnkd.in/gPKXGwv2

You know what feels good on a Friday morning? Closing two of the top 10 most requested community items on your product. ✅ Support timezones in Crons ✅ Support environments without creating deploys You know what feel better? Also shipping multi-label on ARC as we finally can get to what the community asked for there as well 🎉 Being able to come back to the community and just 'engage and fix' what people have been asking for for 4 years is amazing. Next up: �? Parallel steps �?Splitting composite Action steps �?Improve concurrency to not cancel pending jobs And that's just the experience pieces. Watch out for a blog post soon on what we have coming next on the security focused side of Actions 💪

Since its launch, there have been 60 million Copilot code reviews (and counting). 👀 As AI-accelerated development increases the rate of code changes, keeping up with review is getting harder. Copilot code review helps teams close that gap, catching issues early without becoming a bottleneck. If your team is shipping more code faster than ever, this blog digs into how Copilot code review fits into that workflow. 🤖 https://lnkd.in/d5Q5QP73

This week on Open Source Friday, we're talking with Abhi Aiyer, CTO at Mastra, about building AI applications with TypeScript. Mastra is a TypeScript-first framework for creating AI applications. We'll dig into how it works, what problems it solves, and what's next for the project. Links: Mastra GitHub: https://lnkd.in/dVuKrvHu Mastra Website: https://mastra.ai/ About Open Source Friday Every Friday, we highlight open source projects and the people building them.

Choice has always been core to GitHub. Build how you want, where you want, with the right tools, languages, models, and now agents to get the job done. With the rise of models, there’s been a lot of discussion around benchmarks and rankings. They’re useful, but they don’t answer a harder question: how does a model behave inside a real repo, under real revision pressure? We looked at aggregated GitHub Copilot telemetry across 23M+ product requests and examined code survivability, or the percentage of AI-generated code that ultimately makes it into commits. Here are a few observations: 1️�?� Frontier models all show meaningful code survivability. None cluster exclusively in low‑survival buckets. 2️�?� The gap between models is narrower than benchmarks often imply. Across real repositories, most cluster surprisingly close in terms of code that survives. 3️�?� Workflow posture appears to matter more than model choice. Exploratory coding naturally produces lower survival rates than execution-oriented work. Said plainly: when you look at millions of real development sessions, the differences between models flatten more than the discourse suggests. This isn’t about declaring a winner, but rather understanding how models actually behave once they enter the messy reality of real software projects. Long after launch hype fades, this is the signal we care about: does the code survive? Regardless of what the data says, or which model ships next, our role hasn't changed. It’s the same thing it’s always been: making sure developers have access to the best tools. Choice has never been just a feature for us. It’s the whole point (and it informs our whole approach to key tools like Copilot CLI, which lets you set multi-model fleets of subagents loose on any given task). The team’s been cooking up additional insights here. More to come if this is interesting to folks. 👀 The data certainly surprised us. �?�

Come hang and hack with us!

Raycast users: Did you know? 💡 Learn more about what Copilot coding agent can do for you. ⬇️ https://lnkd.in/g2egquxU

Different AI models find different bugs. So why not use all of them? Try this out in Copilot CLI: 1. Run /review 2. Ask it to use multiple model providers at once for a multi-agent code review 3. Get the highest possible signal and catch bugs before anyone else Evan Boyle shows how it's done. ▶️ https://lnkd.in/gqVEVye7

Our first-ever long-term support model is here. Because enterprises shouldn't have to rebuild every time the ground moves. Explore what’s new in the changelog. https://lnkd.in/e8QJiefY

Is the "one size fits all" software model dead? In this episode of the GitHub Podcast, Cassidy and Kedasha explore the rise of personal software tools built by developers for themselves, and shared with the world through open source. https://lnkd.in/gx6ZhTT3