Lab
Visualizations, investigations, and proofs of concept.
Impact Advocate (beta)
An agent skill that turns your own GitHub history into a self-review draft, linked to evidence and calibrated to your role. Runs in your own Claude session. Technical beta, testers wanted.
work-ledger
A local reader for your own Claude Code transcripts. It attributes cost to the sessions, prompts, skills and agents that caused it.
mcp-remote-sudo
Narrow, expiring, receipted authority over a remote Linux machine for an AI agent, instead of an open shell. Early design and prototype.
Tricorder
Turns a repository's code-review history into a map of what a team knows, what it repeats, and where review attention is thin.
Terms & Conscience
Eight AI companies' consumer terms read against a 12-point rubric, and a companion tool that scores any terms of service you paste in.
Seven Grids
San Francisco's street bearings fall into seven sharp peaks — seven surveys that met at awkward angles — and a third of the city that fits none of them.
Five Years of Open Water
244 swims read back out of an Apple Health export that no application had ever been permitted to open. The stroke rate never moved; almost everything else did.
Two Swims, One Route
The same two-kilometre crossing, two weeks apart, at the same stroke rate — and eleven minutes apart. Separating the current from the swimmer using a control that was inside the swim all along.
An Opinionated Guide to Writing Techniques
Ten writing techniques scored against ten organizational genres through ten shared considerations, with the ranking, the selection rule, and the sources behind each one.
Praxis
A deterministic instrument that decides what a message should do for its reader, locates what the draft is missing, and proves the rewrite kept it. It never writes the prose. Looking for testers.
Claude's Built-In Styles
One workplace memo rendered through each of Claude's five built-in writing styles, holding facts, audience and required action constant. A companion to Writing Mechanisms.
Writing Mechanisms
One source statement, six writing mechanisms, and four AI systems—presented as a controlled comparison without scores or rankings.
What ASD-STE100 Is For
Simplified Technical English is circulating as a general standard for good writing. Three models independently reject the claim — STE removes by design the devices persuasion depends on — and the standard body's own training material agrees: it simplifies English for readers, not for writers.
The Standard Before the Tool
Three Claude writing skills measured against the written-communication literature — established first, so the critique could not be reverse-engineered from the thing being judged. The STE skill is the most standards-grounded of the three; nothing in the set tells a writer which technique to reach for.
The Skill Map
An interactive map of 4,975 agent skills across 43 repositories, and the five areas where nobody has published anything yet.
Twine
Pool LinkedIn connection exports from multiple people into a single force-directed network graph. Find your warmest path to anyone at any company.
The Life Expectancy Misconception
Seven interactive views into what 'average life expectancy' actually meant across history — and why the headline number is almost always wrong.
Tab Tamer
A Chrome extension for people with too many tabs — sort, group, search, and bulk-close without leaving the browser.
Captain's Log
A Claude skill that keeps a daily working diary — automatically. Not what you did. What it meant.
Reading With Ears
A personal AI pipeline that turns newsletters into a podcast — three episodes a day, fully automated, from what you were already reading.
Ivy — The Archive
A structured, repo-backed system for capturing and reusing thinking. Your reasoning lives in plain markdown, inside a Git repo — not inside any AI tool.
Evernote Local Vault
A local-first, open-source workflow to archive Evernote notes — keeping everything on your filesystem, searchable, and under your own control.