$ cat ai-agents/week/2026-11-02.log
the week in review · Nov 2 – Nov 8, 2026
Codex pauses Bazel release artifacts for Cargo-only builds
By RepoJournal · composed from the cited sources · human-reviewed weekly · methodology
Codex publisher 0.167.0-alpha.2 ships Cargo assets only, while Gemini CLI cuts v0.67.0 with model and sandbox fixes.
Make Bazel checks advisory and pause Bazel release artifacts (#58717) openai/codex
The BuildBuddy key is revoked, so Bazel-dependent checks no longer block CI and releases publish Cargo artifacts only. If your pipeline pulls Codex artifacts by name, expect the Bazel ones to be missing.
Remove Bazel outputs from DotSlash release configs (#58757) openai/codex
The DotSlash configs depended on Bazel assets absent from the Cargo-only fixture for 0.167.0-alpha.2, so those outputs were removed along with the Windows command runner and sandbox setup entries. Download paths that pointed at them will not resolve.
feat(process-daemon): protocol v8 for shell commands anomalyco/opencode
Protocol v8 for opencode-pty lets OpenCode run its local shell commands through the daemon. Anything speaking the previous daemon protocol has to move with it.
Release v0.67.0 google-gemini/gemini-cli
The 0.67.0 release folds in a dynamic-model fix that keeps an explicit --model gemini-2.5-flash from being overridden, plus a rootless Podman sandbox fix supplying --userns=keep-id for sandbox files created in the bind-mounted workspace.
fix(core): prevent interrupted Whisper model downloads from being treated as installed google-gemini/gemini-cli
Interrupted Whisper downloads previously left a partial file at the final path in ~/.gemini/whisper_models/, so isModelInstalled() reported a broken model as present. Retrying a failed download no longer requires clearing that directory by hand.
fix(cli): serialize AnsiOutput to text content in ACP tool call updates google-gemini/gemini-cli
ACP tool_call_update notifications for terminal output carried empty content arrays when a result's returnDisplay was an AnsiOutput array. In the writer's reproduction, run_shell_command was the path that hit it; shell output now serializes as text.
fix(llms): send reasoning placeholder for models that need interleaved reasoning cline/cline
Reflection Beam rejects requests when an earlier assistant message lacks reasoning_content, so history compacted with dropThinkingBlocks or carried across a model switch returned 400. A placeholder keeps those turns sendable.
feat(desktop): collapse thinking by default and add a setting for it aaif-goose/goose
Desktop starts with the thinking block collapsed, including while the model is working, and a new Show Thinking setting under Settings > Chat (Response Styles card) offers Always, Collapsed or Never.
$ ls ai-agents/week/ # the briefings behind this review
Keep up with Agentic Coding in about 3 minutes: what actually shipped — the commits, pull requests, releases, and security advisories that matter.
fair warning: these emails are deeply technical. diffs, version numbers, CVEs, benchmark deltas. if that's not your idea of a good read, this isn't your newsletter.
Folds into your digest — weekly by default, monthly if you prefer. Unsubscribe in one click.