人工知能

Artificial Intelligence · Applied Sciences · 17 notes.

• "how does <filepath> work?" — Claude Code's placeholder prompt
• A tidy-up of the Claude Agents specifications
• Browser Automation MCP 2026 — Comparing Playwright, chrome-devtools and agent-browser
• Claude Code orchestration glossary, 2026-05
• Claude Code's CLAUDE.md hierarchy — which instructions reach how far
• Claude Mythos and Oceanus — the cutting edge and its "version for everyone"
• Deep Research — auto-generated reports as a new citation touchpoint
• Does Claude Code's Spawn (parallel subagent launch) really run in parallel?
• Evolutionary Model Merge — 進化的探索でモデル合成レシピを自動発見する
• Field glossary — AI-strategy & GEO terms that came up
• Launching Claude Code from an orchestration layer
• Marketing work: compress the analysis, don't fight on research
• MECE prompting — structuring AI output with four letters
• Multi-hop real-task benchmarks — FRAMES / BrowseComp
• Task-type queries — from "questions" to "requests"
• Vector data and vector databases — the foundation of embeddings and semantic search
• What is an LLM (large language model)? — understanding how it works in one go