Daily briefings
AI Briefing: Review APIs, model retirements and Google’s September recap
What changed in Copilot review automation, which models left its selector, and what Google announced during September.
Development
Copilot code review can be requested through APIs
GitHub added REST and GraphQL requests for Copilot code review, including a per-request effort setting. The announcement also confirms that Balanced became the default effort on September 28; explicitly selected Lite settings were retained. GitHub lists the feature as generally available on its paid Copilot plans.
Why it matters
A review can now start from an existing internal workflow. Check effort settings and credit usage before wiring repeated requests into automation.
Source: GitHub · Source published:
Models
Four older models leave GitHub Copilot
GitHub says Gemini 3.5 Flash, Gemini 3.6 Flash, Kimi K2.7 Code and Claude Opus 4.7 were deprecated across Copilot on October 2. Its announcement names suggested replacements and notes that enterprise administrators may need to enable the corresponding model policies. This retirement concerns Copilot availability, rather than every provider’s API.
Why it matters
A saved model choice may stop working even when the surrounding tool remains available. Check integrations and organization policies before assuming a replacement is enabled.
Source: GitHub · Source published:
Industry
Google collects its September AI announcements
Google’s monthly recap brings together model launches, Gemini app changes and research announcements. It covers Gemini 4 Argon, new voice models, Windows access and scientific projects. The roundup is a useful discovery index, but each linked release has its own date and access conditions; the October recap date is not the launch date of every feature.
Why it matters
Use a monthly roundup to find relevant changes, then inspect the original announcement. Availability, eligible plans and rollout timing matter more than the number of launches.
Source: Google · Source published: