Agent Signals

Current state of one tracked coding agent (harness) or model lab: latest verified stable release or recorded models, every finding from the last 30 days with sources, and attributed practitioner takes with stance counts. Use for "catch me up on Pi" or "where is OpenAI at with coding models?". Requires an entity ID from list_tracked.

entity_briefing

Models released (or first detected, when the publisher gave no date) by tracked labs and notable outside publishers, newest first, with publisher, status (announced/available/deprecated/retired), open-weights flag and an official source. Use for "what new models dropped this month?" or "any new open-weight coding models?". Updates to older models do not count as new.

new_models

The weekly Agent Signals briefing: a short summary, highlights and every notable change (importance ≥2) in AI coding agents and models for one ISO week, with sources, practitioner takes and data coverage. Formatted to share with a team. Use for "give me this week's signals briefing". Defaults to the latest completed week.

weekly_briefing

The current Agent Signals watchlist: tracked coding agents (harnesses), model labs and trusted practitioners, with the IDs other tools accept. Call this first when unsure which entity ID to use, or when asked what Agent Signals covers.

list_tracked

Source-cited changes to AI coding agents (Claude Code, Codex, Pi) and coding-relevant models from tracked labs (Anthropic, OpenAI, Google, xAI, DeepSeek, Qwen, Z.ai, Moonshot, others), newest first, each with a source link, a verbatim evidence quote and linked practitioner takes. A change reported by a trusted practitioner before any official source confirms it is labelled as reported, citing their post. Use for questions like "what changed in Codex this week?" or "any new Claude Code features?". Filter by track (harness = coding agents, model = model labs), entity ID (see list_tracked), since, and min_importance (1 minor, 2 notable, 3 changes how people work).

whats_new