Idan Gazit, GitHub, ETLS Vegas (22)
summary
GitHub Copilot workspace inspects a real repo, generates a concrete spec and pinpoints three files to change, then produces an editable PR that boots an in-browser editor so the patch can be tested and tweaked immediately. The claim is that tooling shifts from typing to sense‑making: models do orientation work so reviews match expectations, acting as a second brain for routine changes while still imperfect—this follows a Goldilocks pattern of constrained, actionable suggestions.
AI-written summary · about
related talks
37mGet the Most Out of GitHub Copilot
49mUnlocking Success: Navigating the Code of Life with GitHub CoPilot
17mGitHub Pull Requests for Everyone
40mLevel Up with Copilot
55mKeynote: Exoskeletons, not Autopilots
46mCOPilotING My Autism in Tech
36mME, MYSELF AND AI
22mIncreasing code habitability by using AI coding assistants