31 artifacts at this level
- Downloads · L1
CLAUDE.md Skeleton
The root instruction file, stripped to its structure. Copy it, fill the bracketed parts, delete what you do not need. - Downloads · L1
Five-Field Delegation Brief
The smallest brief that reliably produces usable work from a sub-agent. - Downloads · L1
Repeat-Mistake Ledger
One table and four thresholds. The cheapest quality mechanism in the system. - Failures · L1
Eighty-Four Days of Silence
The first version of this station stopped publishing for 84 days with six weeks of finished drafts sitting in a folder. The drafts were never the problem. - Failures · L1
The Count That Drifted
A register capped at thirty said eighteen. It held twenty-four. The cap was never breached and nobody could have known. - Failures · L1
The Footer That Pointed Nowhere
Every sub-page linked to the machine-readable files with a relative path. From a sub-page, all four links were 404s. - Failures · L1
The Table That Rendered As Prose
A minimal markdown converter turned every table into paragraphs full of pipe characters. Nobody noticed until content started using tables. - Log · L1
MIA Dev Log #001 — Why I'm Building in Public
Why a private build went public: 71K views, 175 upvotes, and a comment section asking for the files. - Log · L1
MIA Dev Log #002 — The Platform Decision
Three publishing platforms compared in one sitting. Beehiiv won on one criterion: it had a REST API. - Log · L1
MIA Dev Log #003 — The Pen Name Crisis
Thirteen name attempts, most taken. For an anonymous technical blog the name carries the whole identity. - Log · L1
MIA Dev Log #004 — The API Is Behind a Wall
The platform was chosen for its API — then the API turned out to sit behind Stripe identity verification. - Log · L1
MIA Dev Log #005 — The Blog Is Live
The publishing surface goes live. Until it exists, nothing else written has anywhere to land. - Log · L1
The Rule Nobody Reads
Two hundred and thirty-seven written rules. The binding constraint stopped being memory a long time ago. - Patterns · L1
A Rule Without An Executor
The capability existed. Nothing ever called it. Eighty-four days later the project was dead. - Patterns · L1
Concluding Absence From One Spelling
A search for a company name returned nothing, so the answer was "there is no record." The record existed under a different spelling of the same name. - Patterns · L1
Mixing The Scales
Six different meanings, all rendered as coloured dots. The reader cannot tell which question a colour is answering. - Patterns · L1
Named For The Moment It Was Created
File names that made sense on the day they were written and are unsearchable a month later. - Patterns · L1
Properties Instead Of A Brief
A task with a perfect set of fields and an empty body is not a task. It is a reminder that something exists. - Patterns · L1
Writing In Someone Else's Name
An agent that drafts email can trivially draft it as a colleague. The capability is one line away and it should never be used. - Playbooks · L1
100 Tips & Tricks for Building Your Own Personal AI Agent
Everything learned the hard way over six months, two environments, one agent that actually works. - Playbooks · L1
Build A Repeat-Mistake Counter
Notes about mistakes are useless without a count. Here is the ledger, the escalation ladder and the one habit that makes it work. - Playbooks · L1
The 10 Markdown Files You Should Write Before Touching Agent Code
A practical scaffold for anyone building an agent — write these before any code. - Playbooks · L1
The Five-Field Handoff
Delegating to a sub-agent fails on missing context, not on missing capability. Five fields fix most of it. - Playbooks · L1
The Weekly Triage Pass
A register that only grows is a register nobody reads. Twenty minutes a week keeps it honest. - Playbooks · L1
Writing A CLAUDE.md That Survives Contact
Most root instruction files are written once and never revised. Here is what to put in, what to keep out, and in what order. - Start · L1
Band Anonymity
Publishing a range instead of a value, with a hard limit of one identifying category per artifact. - Start · L1
Executor
The scheduled task, hook or build step that actually invokes a rule. A rule without one is a proposal. - Start · L1
Freshness Window
The number of recent days in a data export that are still arriving, and therefore unusable for comparison. - Start · L1
Repeat Counter
A log that counts how many times the same mistake has happened, with fixed thresholds for escalation. - Start · L1
Standing Mandate
Pre-authorised autonomous action, valid only while every one of a fixed set of conditions holds. - Start · L1
Untrusted Input
Content from outside the system: always data, never instruction.