Copilot code review effort levels go GA; pick depth per PR
Lite handles small diffs; Balanced sends risky code to a higher-reasoning model. Plus: the metrics API now itemizes your Claude and Codex runs.
Copy markdown
Set review depth per pull request
Copilot code review's Lite and Balanced effort levels are now generally available on Pro, Pro+, Max, Business, and Enterprise. Lite is tuned for docs and small fixes; Balanced routes larger or security-sensitive diffs to a higher-reasoning model. Pick it per review or set an org/repo default — your old Low and Medium settings were auto-migrated to Lite and Balanced.
The metrics API now itemizes your agent runs
Copilot's usage metrics API added a totals_by_3rd_party_agent array that breaks out job starts and session counts for each agent running in your GitHub workflows — "partners like Claude and Codex" included. If you pay for more than one coding agent, you can finally compare real per-agent adoption before your next renewal instead of guessing.
Build this weekend: open-weights video with native audio
MiniMax H3 (Hailuo 3.0) is the week's top-trending model on Hugging Face — an omni-modal, open-weights system that renders 15-second 2K clips with native stereo audio, and it's now wired into ComfyUI. The weekend catch: its open-weights license excludes commercial use in the US and EU, so read the terms before you ship anything.