Each entry covers a production release: what we shipped, what changed for you, and anything you need to know if you build on the API or MCP server.
Prompt runs now survive restarts and deploys without dropping or duplicating prompts, and recover on their own when stranded. Plus more accurate brand detection with per-mention alias matching, editable brand aliases in project settings, and more complete answer coverage.
Tag prompts with branded/non-branded and custom labels, track goal-oriented workstreams over time, and a product-wide rename from keywords/queries to topics/prompts. Includes the May 8 release: Grok and Gemini coverage, more accurate brand detection, and public docs.