James Brady — builds AI systems that show their work

01Proof

4 things you can go look at right now.

Each one links to the thing itself: a repo, a doc site, a running product, and not a case study about it. Every value and every last-active date is read from the content source when the page is built.

OfOne Active

of 90 predeclared benchmark slots run

OfOne is an open-source decision compiler that turns a hard question into a typed, checkable map before it writes any prose.

Source: Public repo · docs site
Last active: 2026-08-01
Go: Read the docs

plimsoll Active

$48.46

one merged PR, measured end to end

Plimsoll is a local-first collector that ties AI coding spend to shipped outcomes, so a team can say what its tokens actually bought.

Source: Public repo README
Last active: 2026-07-20
Go: See the method

The visibility platform Live

measures under 12 fixed axes

The visibility platform measures how findable a business is in ordinary search and inside AI assistants, then scores it against a fixed rubric of 12 axes and 76 measures.

Source: Published method site
Last active: 2026-08-11
Go: How it's measured

EEG meditation toolkit Dormant

stars from outside this project

The EEG meditation toolkit is Python tooling that turns a raw consumer-headset recording into a readable account of a meditation session.

Source: github.com/CryptoJym
Last active: 2025-11-11
Go: View on GitHub

02Work

Products, open source, client work, experiments.

Filter the shelf. The count below updates with it. The number is read from what is on screen, not typed in, and there is no denominator to fall out of date.

Products

Decision compiler

OfOne

An open-source decision compiler: a hard question becomes a typed, validated map, and the prose you read is a rendering of that map.

Docs live · MIT

Open source

Spend telemetry

plimsoll

A local-first collector that joins AI coding spend to merged pull requests, so cost per shipped outcome stops being a guess.

Unit: dollars per merged pull request

Products

Findability, measured

The visibility platform

A multi-tenant platform that measures how findable a business is in search and in AI answers, and prints the evidence under every score.

In production · Anonymized

Open source

Signals, not vibes

EEG meditation toolkit

Python tooling that filters an EEG recording, measures band power, and scores meditation depth. It runs without hardware.

Python 3.8+ · 3 stars

Products

Marketing site that has to prove itself

seopr1.com

A five-page site selling AI-search visibility, built speed-first: static pages, one dependency, and an explicit fallback for every effect.

Live · 1 production dependency

Products

A shared frame, in one sitting

AI-readiness assessment

An interactive assessment scoring seven AI-competency domains across 28 role-filtered questions, with four views of the same result.

7 domains · 28 questions

Experiments

A thesis in progress

The Of One family

A governed family of role-shaped domains for one-person businesses: two live applications, the rest deliberately staked ground.

Live apps vs staked ground — labelled

Showing Filter:

03Theories

Open questions I'm working in public.

Labeled by what they actually are. Anything still named-only stays off this page until it is at least sketched.

T-01

Universal Question Geometry

Flagship · working paper

Questions have shape, and the shape predicts what an answer can reach. A map of how questions sit next to each other, and why some open doors that others structurally cannot.

ActiveRead it

T-02

Question–Answer Dynamics

In daily use

What an answer does to the next question. The loop, measured: most of the value in a conversation lives in how the second question moved.

ActiveRead it

T-03

The Architect Loop

Field-proven · fork, not mine

One model holds judgment, a separate set of models does all the building, and a git repository is the only shared memory. A fork of Dan McInerney's project, hardened and run daily.

ActiveRead it

T-04

Latent Emotions in LLMs

Paused exploration — live demo

Different models reach for closely similar words when describing certain relational states. That might point to shared structure underneath, or to nothing. The work is parked, and it says so.

PausedRead it

T-05

THE PAPER

Introduction only — document unpublished

A specification for running a business where AI agents do most of the labour, in five gated layers. The document itself stays private; this page introduces the idea and links nothing.

ActiveRead it

T-06

Movement Economy

Sketched · enforced in code

Every sentence in a piece of analysis must do one of ten declared jobs or be deleted. Enforced on map objects inside OfOne rather than applied by taste to prose.

ActiveRead it

T-07

Function-First Orchestration

Sketched · standing rule

Ship one honest happy-path test plus the refusals that protect money, custody and truth. Defer the rest deliberately, in writing, rather than skipping it.

ActiveRead it