CLAUDE LABJP
CLI — Claude Code v2.1.259 landed on September 2. It is a dense release focused on org-wide configuration, unattended runs, and permissions — the most practically useful one this weekMCP — A new managedMcpServers setting lets an organization push HTTP/SSE MCP servers to every user. At the same time allowedMcpServers narrowed in scope, so an allowlist no longer filters managed servers outHEADLESS — The new --permission-prompts none flag denies anything that would prompt, while the active permission mode, auto included, keeps making its own calls. Useful for anything you run unattendedSECURITY — Bash Read() deny rules had gaps: files passed as option values such as --ignore-revs-file=.env, and compounds like cd DIR && cat FILE, slipped past them. Both are now coveredSLACK — Claude Tag no longer judges messages one at a time. It reads the whole channel, along with its memory and your standing instructions, and is roughly 30% better at staying out of conversationsLEARN — Claude Academy is open. It is a free learning hub built on the 4D AI Fluency Framework that Anthropic uses to onboard its own team, with completion tracked in your Claude profileCLI — Claude Code v2.1.259 landed on September 2. It is a dense release focused on org-wide configuration, unattended runs, and permissions — the most practically useful one this weekMCP — A new managedMcpServers setting lets an organization push HTTP/SSE MCP servers to every user. At the same time allowedMcpServers narrowed in scope, so an allowlist no longer filters managed servers outHEADLESS — The new --permission-prompts none flag denies anything that would prompt, while the active permission mode, auto included, keeps making its own calls. Useful for anything you run unattendedSECURITY — Bash Read() deny rules had gaps: files passed as option values such as --ignore-revs-file=.env, and compounds like cd DIR && cat FILE, slipped past them. Both are now coveredSLACK — Claude Tag no longer judges messages one at a time. It reads the whole channel, along with its memory and your standing instructions, and is roughly 30% better at staying out of conversationsLEARN — Claude Academy is open. It is a free learning hub built on the 4D AI Fluency Framework that Anthropic uses to onboard its own team, with completion tracked in your Claude profile
TAG

groundedness

1 articles
Back to all tags
Related:
Claude API1RAG1vector search1retrieval augmented generation1evaluation1TypeScript1
API & SDK/2026-07-09Advanced

When the RAG Started Being Confidently Wrong — Field Notes on Measuring Retrieval Misses With Groundedness

In a Claude API RAG, the answers stay fluent while the facts drift. Often the cause is a silent recall decay on the retrieval side, missing the document that holds the answer. Field notes on measuring groundedness and retrieval hit rate and walking the system back, with working code and real numbers.