$ cat ai-agents/week/2026-11-16.log
the week in review · Nov 16 – Nov 22, 2026
Claude Code 2.1.326 makes Code Mode default for agent teams
By RepoJournal · composed from the cited sources · human-reviewed weekly · methodology
Codex 0.170.0 adds realtime voice in the TUI and codex exec gains a --remote flag; Gemini CLI 0.69.0 stops shell expansions in templates.
v2.1.326 anthropics/claude-code
by claude-code-publisher[bot]
Code Mode now turns on for agent-team sessions you had already enabled it for, so the startup path for team runs changes without you opting in again. The same release adds a startup warning for --allowedTools, --disallowedTools and --tools entries that don't look like a complete tool name, which catches typos that previously failed silently.
CLI v3.0.75 cline/cline
by github-actions[bot]
A prompt argument or piped stdin now starts in yolo mode rather than act mode, since nobody is at the keyboard to approve tools. Act mode is preserved when auto-approve is off, when a mode flag is passed, when plan mode is saved, or when the run is non-one-shot, so any CI job relying on the old default needs a flag to keep its approvals.
0.170.0 openai/codex
by github-actions[bot]
The TUI gets WebRTC negotiation for realtime voice calls, and OpenTelemetry HTTP exporters now refresh authentication headers automatically. MCP servers with the same name across different environments are also supported experimentally, which matters if you run one config against several backends.
Support remote app servers in `codex exec` openai/codex
codex exec --remote <ADDR> runs a noninteractive turn using an existing app server's configuration and credentials, starting a new thread or resuming one by UUID, with human-readable or JSON output and optional bearer-token auth. Long scripted runs can now reuse a server you already have running instead of re-authenticating per invocation.
Release v0.69.0 google-gemini/gemini-cli
Template placeholder replacement now preserves literal $ characters, so prompts containing shell syntax survive intact instead of being expanded. The release also adds a /stats heatmap subcommand for daily token usage, and IDE companion discovery files are scoped to a per-user temp directory with 0700 permissions and written atomically.
fix(server): answer handler defects with a readable 500 instead of an empty one anomalyco/opencode
A crash in a handler used to return a 500 with an empty body. That 500 now carries JSON: what went wrong with storage where that's the cause, otherwise a reference for finding the crash in the server log, with messages that never name the database engine.
fix(cli): skip system MCP servers with insecure command permissions google-gemini/gemini-cli
Local MCP server commands and script arguments referenced from settings.json and system-defaults.json are now checked for ownership and write permissions, so a system-wide config can no longer point at a command you don't control. Worth knowing if you distribute MCP setup through system defaults.
remove chat mode (#13571) aaif-goose/goose
Alongside: Cline's SDK v0.2.1 stops re-cutting read_files results to the 8k tool-result cap so a full read arrives in full, Goose 1.60.x drops chat mode and bounds empty-response recovery to three retries per InferenceRunner step, and opencode adds mid-conversation tool changes and automations that wake a session only when there's something to act on.
$ 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.