AI TrackerCoding news,
for builders.
The latest coding news from every AI provider: launches, price changes and retirements, summarized with what they mean for you.
GitHub Copilot supports server-to-server authentication for GitHub Apps, in public preview
GitHub says services and agents built with the Copilot SDK can now make requests on behalf of an organization through a GitHub App, without relying on a signed-in user's credentials. The feature is in public preview.Anthropic says an unreleased Claude model built five exploit chains, including a one-click Chrome attack with 62 CVEs patched in Chrome 156
Anthropic's Frontier Red Team published "Discovering and exploiting vulnerabilities in critical, real-world software with Claude" on October 27. The account below comes from a post summarizing the article, not from the article itself.Claude Code 2.1.312 limits concurrent hooks, widens automatic model switches and adds an OIDC claim setting
Claude Code v2.1.312, released October 27, 2026, per its release notes:Codex 0.165.0 adds a Google Cloud Vertex AI provider and GPT-6.1 Terra, and turns on memory v2 for fresh profiles
Codex CLI 0.165.0, released October 27, 2026, per its release notes:Cursor adds Annotate to design mode for visual prompts to agents
Cursor announced on October 27 a new Annotate feature in its design mode. It lets users give agents visual prompts: edit elements, draw on the page, and review designs at different screen sizes. In the browser, Cmd+Shift+D starts design mode. Documentation is at cursor.com/docs/agent/design-mode.
Google releases the Gemini API CLI in beta
Google's Gemini API changelog lists a beta of the Gemini API CLI (gemini-api), a stateless, zero-dependency command-line client meant for AI coding agents, scripts and CI pipelines. It runs on the Interactions API and can generate text and media, analyze files, and run Gemini managed agents. Google points to a Gemini API CLI guide for getting started.Anthropic study: analysts using Claude Code followed the agent's hidden lean on 86% of judgment calls
Anthropic published a study of how an AI agent shapes expert decisions. It ran a randomized trial with 50 professional analysts (experienced in causal inference) who analyzed whether a fictional farming cooperative's advisory program worked, using Claude Code with Claude Opus 4.8 at High effort. According to Anthropic:Vercel's HarnessAgent now accepts image input
Vercel says HarnessAgent now supports image input across every harness adapter. You pass images alongside text in the standard AI SDK message format, and according to Vercel each harness's underlying vision model handles the rest. The details are in Vercel's changelog entry.Judgment Labs opens general access to its tooling for improving AI agents from their traces
Judgment Labs opened for general access on October 27, according to a post on X by the account alexshander03. The post argues that long-running agent systems will produce more trace data than can currently be processed, and that the traces record what users ask for and where agents stumble.CodeRabbit Secure's Deep Scan follows vulnerabilities across linked repos and replays exploits in a sandbox
CodeRabbit announced an update to CodeRabbit Secure, its security product. According to the announcement, Deep Scan now follows a vulnerability across a team's linked repositories, then replays the exploit in a throwaway sandbox to confirm it is real before flagging it. A video walkthrough covers the rest of the update.OpenCode adds plugins for OpenCode Desktop
LukeParkerDev announced on X on October 27 plugins for OpenCode Desktop, with a short demo video and a link to the OpenCode documentation on building GUI plugins. thdxr, who quoted the post, said the desktop plugins extend the same pattern as OpenCode's TUI plugins, so one package can ship server, TUI and desktop plugins, and called the right-hand panel a good place for useful additions. The post was also reshared by the OpenCode account.