Blog
Essays and guides on shared AI context — context collapse, convention drift, and making every AI tool your team uses arrive briefed.
An ESA prime ships its AI context by copy-paste
Starion publishes team-shareable Claude Code skills. The documented install path is cp -r. Why the copy step is where shared AI context drifts.
Read post
Your AI context is now a supply-chain attack surface
The keyv npm worm's second execution path was .claude/settings.json — no install required. Why shared agent context is unsigned executable config.
Read post
Who actually has team AI memory? A dated comparison
One question sorts every vendor: can memory one person accumulates be read by a colleague? Answered per vendor with dated primary sources, August 2026.
Read post
AI Context Lock-In: The Five Tests It Has to Pass
Inspection, Export, Replay, Audit, Deletion — the five-test rubric for AI context lock-in, run honestly against OpenAI, Gemini, Anthropic and us.
Read post
Should you delete your CLAUDE.md for Opus 5?
Anthropic cut 80% of Claude Code's system prompt and flipped CLAUDE.md to auto-memory. What to delete, what to keep, and why 405 comments pushed back.
Read post
MBSE Shared Model vs Your AI Copilots' Context
MBSE gave aerospace teams one authoritative system model. But engineers and AI copilots still run on scattered context. Here's the layer that closes the gap.
Read post
Context rot: why 1M-token windows make agents dumber
Context rot is the decline in an AI agent's reliability as its window fills with low-relevance tokens. Why 1M-token models make it worse — and the fix.
Read post
Personal AI memory is free. Team memory is the hard part
Copilot Memory and open-source layers made personal AI memory a free commodity. Governed team memory — roles, audit, portability — is the unsolved part.
Read post
How to Share Cursor Rules Across Your Whole Team
Cursor rules stuck on one laptop? Store them once in a shared workspace so every developer's Cursor — plus Claude and Copilot — stays in sync. Here's how.
Read post
MCP Just Went Stateless: The 2026-07-28 Spec Explained
MCP's 2026-07-28 spec drops the session and goes stateless. Here's what changed, what breaks, and where your team's shared context should live now.
Read post
AGENTS.md vs CLAUDE.md vs SKILL.md: Which to Use?
AGENTS.md, CLAUDE.md and SKILL.md compared — three altitudes of one instinct, why all of them drift at team scale, and the layer that actually fixes it.
Read post
Aerospace tribal knowledge: shared AI context for teams
Aerospace runs on tribal knowledge against a retiring workforce. Learn why AI tools start blank and how a shared context layer turns it institutional.
Read post
The CLAUDE.md problem — why per-developer context files fail at team scale
CLAUDE.md files are the right instinct and the wrong architecture for teams. Where per-developer context files break down — one tool, one repo, one person, one size ceiling — and what a fix has to look like.
Read post
How to give Claude Code persistent context across sessions
A practical guide to making Claude Code remember your project: CLAUDE.md done properly, user-level instructions, auto memory, and MCP servers — plus when each layer stops being enough.
Read post
The best AI memory and context tools in 2026 — an honest comparison
CLAUDE.md, Cursor Rules, Mem0, Pieces, Claude and ChatGPT Projects, and Contextium compared honestly: what each actually is, what it costs, where it breaks, and which to pick for your situation.
Read post