GitHub Copilot: Coding Will Never Be The Same Again
summary
GitHub Copilot powers code completion with GPT‑3.5‑turbo for ~300–400ms completions and GPT‑4 for chat, deletes prompt context, blocks suggestions over ~150 characters (≈60 lexems) and scans GitHub to avoid license or PII leaks. Product stats: 1.5M developers, 37k organizations. Studies: a 6‑month cohort of 450 developers shows 50% more PRs, 15% more merged PRs and 45% higher build success; earlier lab studies cut task time roughly in half. Demoed features include workspace embeddings, R.A.G. over org repos, conversational fixes and automated refactors.
AI-written summary · about
chapters (14)
- 0:00Intro
- 1:10Software has already eaten the world
- 5:38Demo
- 11:55What's happening under the hood?
- 21:41Demo
- 29:11Retrieval augmented generation (R.A.G.)
- 30:38Demo
- 32:45R.A.G. continued
- 34:40Custom models
- 39:36Research
- 45:326-month study of 450 devs
- 47:38Global impact
- 48:56Syntax is not software
- 50:15Outro
related talks
01h 00mLevel up with GitHub Copilot: using AI to learn, code, and build
40mUnlocking Productivity with GitHub Copilot Best Practices and Tips
01h 20mGithub Copilot - The World's Most Widely Adopted AI Developer Tool
57mAccelerate Your Developer Productivity with AI Embrace the Future Now!
46mPragmatic techniques to get the most out of GitHub Copilot
25mImproving developer happiness with AI
45mSupercharge your coding: Integrating ChatGPT and Copilot in long-term projects
01h 00mGitHub Copilot, using AI to help you learn, code, and build