Your Box and Your Trust Model
What runs on your machine, and how much rope you give the agent — one has a project answer, the other is yours
Topic
AI
24 cairns tagged "ai"
Working in Parallel (Mostly)
Worktrees, the bd worktree warning, and the one Docker constraint we have not solved yet
Timbers, the Ledger
The second ledger every commit needs — one line of what, why, and how, searchable forever
The Workshop
A map of every tool between you and a shipped pull request, and where each one comes from
Quiet Power Tools
RTK, tokf, and the small CLI multipliers that compound across every session
Quality Gates: The Contract That Lets You Move Fast
Lint, types, tests, coverage, audits — the deterministic constraints that make agent-driven work safe to ship
Just: One Place to Discover
The discovery surface that turns project incantations into a known place humans and agents both look first
How We Build Here
What's changed about writing production software, and what this trail is going to teach you
From Plan to Pull Request: A Day on Strike
One realistic feature, end to end, with every prior cairn put to work
Codex as Second Opinion
When a single model is bluffing or stuck, a second model is the cheapest unblock available
Claude Code as Daily Driver
The agent harness most of us reach for first, what we run inside it, and the alternatives that are perfectly fine
Beads, the Backbone
The CLI issue tracker that gives an agent a memory that survives the next compaction
The Work That Writes Itself
Why we built Conduit, and what it gives the team that nothing else did
How Conduit Runs
Onboarding a project, surviving operational failure, and what's deferred until the system grows
How Conduit is Shaped
The architecture that lets one consumer site read N project repos without coupling either side to the other
From Intake Folder to Project Memory
A design pack for turning KCCI project folders into a managed, searchable, auditable knowledge system
Three Memories, One Q
Why the current OpenClaw design keeps long-term memory private to each lane and shares knowledge through docs instead
The Only Locked Door
How OpenClaw gives Q a real computer without letting public channels rewrite the machine it runs on
Surviving the Upgrade
What happens when the ground shifts under a running AI agent — and how three days of misdiagnosis led to a five-minute fix
Timbers
A Development Ledger for the Age of AI Agents
The Weekend Knowledge Base
Building a self-improving knowledge system from zero in under 8 hours of actual effort — and what that says about internal tooling in 2026
The Memory Problem
Why AI Agents Forget Everything and What To Do About It
The Injection Problem
Defending LLM agents against prompt injection when they read the world
The Quiet Teammate
What Happens When an AI Agent Picks Up the Work Nobody Owns