Blue Moose logo Blue Moose Field notes
Tag / AI
Claude Code's October release run tightened the blast radius: workspace trust, sandboxed writes, and a fast-mode sunset
claude code

Claude Code's October release run tightened the blast radius: workspace trust, sandboxed writes, and a fast-mode sunset

Three releases in nine days — v2.1.309, v2.1.314, v2.1.316 — added an explicit trust gate on ultrareview, a setting that refuses file writes outside the working directory in every permission mode, idle-compaction for long-running sessions, and a deprecation notice for Opus 4.8 fast mode. None of it is a new feature; all of it is about what the tool is and isn't allowed to touch.

BM Blue Moose / 2026-11-02 / 03 min
2026-11-02 / 03 min
Claude Code shipped the per-subagent effort control my own config had been pretending existed
ai

Claude Code shipped the per-subagent effort control my own config had been pretending existed

I write subagent model-selection rules into my global CLAUDE.md as a matter of habit. A string of October releases quietly added the actual knob for it — a per-agent effort parameter, fork subagents on by default, and the UI finally showing the effort level actually in force. Here's the commit history, from someone who uses this daily.

BM Blue Moose / 2026-10-28 / 03 min
2026-10-28 / 03 min
Does my CLAUDE.md actually matter? Reading the 'A Few Pages of Markdown' paper
ai

Does my CLAUDE.md actually matter? Reading the 'A Few Pages of Markdown' paper

An ASE '26 paper reports that repos with committed AI config show about half the complexity growth after adopting coding agents. Tempting to read as 'write a CLAUDE.md, halve your mess' — but it's observational and the authors call it hypothesis-generating. A skeptical read, and the audit it actually suggests.

BM Blue Moose / 2026-10-22 / 04 min
2026-10-22 / 04 min
Craft 6 puts an MCP server in core. A Drupal dev reads the permissions screen first.
craft cms

Craft 6 puts an MCP server in core. A Drupal dev reads the permissions screen first.

The Craft 6 alpha ships built-in Model Context Protocol servers that let an AI inspect and edit content, the content model and project config. The interesting part isn't the AI — it's the permissions design: OAuth 2.1, an explicit 'Use Craft MCP' permission, public access off by default, and writes that land as drafts.

BM Blue Moose / 2026-10-19 / 04 min
2026-10-19 / 04 min
Agent Skills are just folders of Markdown — and after a year of MCP hype, that's the point
ai

Agent Skills are just folders of Markdown — and after a year of MCP hype, that's the point

The most useful upgrade to my agent workflow lately is embarrassingly low-tech: a folder with a SKILL.md in it. Anthropic's Skills (now an open standard) use progressive disclosure to load the right instructions only when relevant. They're the honest counterweight to MCP: half the time you don't need a live tool, you need the paragraph the agent was missing — versioned in git, no daemon. Plus the supply-chain caveat nobody screenshots.

BM Blue Moose / 2026-09-07 / 04 min
2026-09-07 / 04 min
RAG cost per query is a lie — the real bill is the re-index you haven't done yet
ai

RAG cost per query is a lie — the real bill is the re-index you haven't done yet

Everyone budgets RAG by the per-query LLM call, then gets blindsided three sprints in when they change chunking or swap embedding models and have to re-embed the whole corpus. Real numbers, why an embedding model is a lock-in decision disguised as a benchmark score (incompatible vector spaces), and how to budget for the re-index — plus when it genuinely doesn't matter.

BM Blue Moose / 2026-09-02 / 04 min
2026-09-02 / 04 min
WebMCP: your website becomes a tool the browser agent calls — read the fine print
ai

WebMCP: your website becomes a tool the browser agent calls — read the fine print

A Google/Microsoft-backed proposal lets your page hand an AI agent real callable tools instead of screenshot-and-guess. Chrome's already trialing it. But it's a W3C draft (not a standard), the API is churning, and the security fine print is stark: tools run with your logged-in auth and there's no mandated consent model. Clever, and a new attack surface — both at once.

BM Blue Moose / 2026-08-24 / 04 min
2026-08-24 / 04 min
Subscribe to AI