GitHub

Better tools didn't improve Copilot code review. Rewriting their instructions did. We swapped Copilot code review onto the same shared tools that power the Copilot CLI (grep, glob, view), expecting a clean upgrade. Instead, reviews did more work and caught fewer issues. 👀 The tools were fine. Their instructions were written for a different job, so the agent browsed the repo instead of investigating the pull request. We rewrote them to mirror how a reviewer actually reads a pull request: ask, narrow, read, decide. The result: roughly 20% lower average review cost, while maintaining the same review quality. ✅ The bigger lesson: for an agent, tool instructions are as load-bearing as your API docs. Here's how we traced the regression to the instructions ⬇️ https://lnkd.in/eDeJg8Gy

What does open source maintenance actually look like once a project starts to scale? In this Open Source Friday stream, Ayushman Bhattacharya shares how Pollinations.ai built the systems behind a fast-growing open source AI project. We’ll dig into contributor workflows, paid quests, AI-assisted PR review, app submission pipelines, model curation, and what maintainers really spend their day doing.

We have a new dataset for your next build, and it's open source. 📊 This time, we're researching non-English natural-language content. Our GitHub Multilingual Repositories Dataset spans 40M+ repos and 80M+ classification rows, showing where non-English READMEs, issues, and PRs live. Some highlights: • Korean is the most common non-English language in issue text, but only the fifth-most common in READMEs. • Portuguese tops the non-English README list with more than 3 million repositories. Find out how you can use these metrics to inform your next project 👇 https://lnkd.in/eRYj7w6T

The automations feature of the GitHub Copilot app is one of the most powerful tools I've got. I have a bunch of automations now - but here's an example of one that triages my inbox and automatically takes action. Here's how it works: The automation fires every day at 9 AM. It goes through my inbox and categorizes things as... * Notification * Trash * Archive * Reply Then it assigns each a confidence score. If the confidence is above 95%, it takes action without waiting for me. Replies always have to be approved. Then I tell it after each run what it got right and wrong and it corrects, updates a learnings.txt file and then applies those learnings on the next run. I'm now to the point where the agent pretty much gets it right every time - took about 2 weeks of training. This is what I want - agents that can handle things autonomously and only bother me when I need to know about something. I blurred the bank balance so you won't be depressed on my behalf.

Let's talk about the latest GitHub updates and cowork together!

📣 OpenAI's GPT-5.6 family is now generally available and rolling out in GitHub Copilot. GPT-5.6 comes in three variants: ☀️ GPT-5.6 Sol: the highest reasoning ceiling in the family. Best for complex reasoning over large codebases and demanding, long-running agentic work. 🌎 GPT-5.6 Terra: the balanced default. A strong all-round choice for everyday interactive and agentic coding. 🌕 GPT-5.6 Luna: a lightweight, cost-efficient variant for smaller, faster tasks and the lowest-cost option in the family. Try one, two, or them all in Visual Studio Code or the GitHub Copilot app. https://lnkd.in/e7MMxgCE

This is the last time we'll mention Super Early Bird. Mostly because it ends today. Get your $600 discount. We'll see you at GitHub Universe.👇 https://lnkd.in/eTPMeFDu

When a pipeline fails, nobody knows until users start complaining. On July 15th, we're joining forces with Andrea Griffiths, Senior Developer Advocate at GitHub, to show you how to build a loop that fixes that. Sentry catches the error, Seer figures out why, Copilot writes the fix, and a pull request gets opened automatically. We're breaking a real script live so you can see exactly how it works. Learn how to set up the same flow in your own repo. Register here: https://lnkd.in/gftMja3S

Have you explored the new GitHub Copilot app? Join our beginner-friendly Let's Learn sessions to get started. With the app, you can: • use AI agents to explore issues, build features, and review changes in parallel, without switching between tools. • set up isolated local and cloud sessions to investigate bugs, implement features, respond to feedback, and experiment with changes simultaneously without conflicts or interruptions. Register for your session in English, Korean, Chinese, Japanese, Spanish, or Portuguese. 🗓️ https://gh.io/letslearn

The GitHub Copilot app is now available on every Copilot plan. 🎉 No plan? You can still build, just bring your own key. 🔑 Here's how to set it up. Download the app 👇 https://lnkd.in/eyS6auvP