Notes on agents, memory and handoffs

What we're learning about running AI agents on real projects: shared memory, decisions, handoffs and the market around them.

• 29 Oct 2026·7 min readclaude-codeagent-memoryhandoff

  Keep project context across Claude Code sessions

  Every new session starts at zero. The fix isn't a bigger context window — it's a record the agent reads before its first prompt.

• 28 Oct 2026·7 min readclaude-mdagents-mdcursorconventions

  CLAUDE.md, AGENTS.md and .cursor/rules: why they rot, and what belongs somewhere else

  Keep stable conventions in repository instructions. Move changing state into reviewable handoffs, and distinguish human decisions from observed work and proposed notes.

• 27 Oct 2026·8 min readcursorclaude-codecodexswitching-tools

  How to switch from Cursor to Claude Code (or Codex) mid-project without losing your decisions

  Switch from Cursor to Claude Code or Codex with a reviewed handoff that separates human decisions, observed work, and proposed next steps, then verify it before editing.

• 26 Oct 2026·7 min readhandoffcontractorsagenciesclients

  How to hand a project to a contractor (or a client) without a two-hour call

  Hand off a project with clear scope, verified access, evidence-backed status, decision rationale, and a bounded next action, without turning the brief into a long call.

• 25 Oct 2026·8 min readdecisionsclaude-codeguardrails

  Why your agent keeps "fixing" code back to the old way — and what actually stops it

  Stop accidental reversals by recording approved decisions with reasons and scope, separating observations from proposals, and checking edits against current constraints.

• 23 Oct 2026·7 min readagent-teamsinstinctmusemulti-agent

  Running several AI agents? Give them one shared memory — and keep yourself as the only one who decides

  Give agents a shared brief without turning suggestions into decisions. A practical workflow for separating human approvals, observed work, and proposed notes.

• 22 Oct 2026·7 min readcomparisonmem0basic-memoryentireagent-memory

  Agent memory tools, honestly compared: memory APIs, session capture, notes, and a shared record

  Compare agent memory approaches by their job: user recall, session evidence, persistent instructions, or a shared project record. Then test what survives a handoff and who approves it.