Modelwright Studio
Humane systems for complicated minds.
A modelwright builds the working arrangement between a person and their machine intelligence — the files, the roles, the boundaries, the habits — wrought to fit one particular human life.
The craft
A wheelwright built wheels. A shipwright built ships. A playwright — the old spelling tells the truth — wrought plays: worked them like material, joint by joint, until they carried weight. A modelwright builds the working arrangement between a person and their machine intelligence.
The craft rests on one old idea and one new problem. The old idea: your mind has never been entirely in your head. A notebook, a calendar, a filing cabinet — philosophers call this the extended mind (Andy Clark and David Chalmers named it in 1998), and every working person has always known it by feel. Your tools think with you. When they're well-arranged, you are smarter than yourself. When they're a mess, you are dumber than yourself, and you feel it, and it costs you.
The new problem: machine intelligence joined the extended mind without an invitation or a job description. Now the notebook talks back, and the arrangement between a person and their tools has become a genuine problem of fit — of joinery, of what answers to what.
Modelwright Studio watches advanced agentic computing where it first emerges, then translates it into humane, portable, understandable practices before it becomes an opaque consumer default.
The commitments
The system fits the person
Not the person the productivity industry wishes you were — the one who actually shows up, with that mind, that energy, those projects, that history of abandoned setups.
The human governs
Judgment, values, authorship, and final say remain with the person. A system that quietly replaces your judgment hasn't extended your mind; it has evicted you from it.
The build outlives the builder
Everything is documented in plain files, readable by any model and by the person themselves. A system is finished when it can run without the wright in the room.
No provider owns your system
The wright runs six different engines behind one dispatcher, daily, swapping between them by what's free, fast, or right for the task at hand — the same portability a build gives you. Files and prompts built to run through a command line, not a vendor's login screen, outlast any one company's roadmap: when a model is deprecated, the system typically moves to another one under routine stewardship — a small change, not a rebuild. The rare exception gets scoped and quoted in writing before it starts, the same way any larger piece of work would. "Any model" means any model you already have access to, not a promise of unlimited reach.
The offers
Published prices — hiding them just re-creates the sales call.
00 · Workbench Hours $75 – $350
One structured 60-minute conversation about a single live question — which tool to use, how to unstick one workflow, what's actually wrong with one part of your system. Not a full audit: real, focused thinking-through, plus a short written follow-up within 48 hours. Friend and community rate $75–100; standard $150–200; business or organizational $250–350, routed by one question at booking rather than a guess about what you can afford. One session, one follow-up — and if it's worth doing again, it's priced the same every time. This isn't the road to a build; that's the Diagnostic below. It's for when you have one question and want one good answer. Email me the question — no application needed.
01 · The Diagnostic $0 founding rate / $200 standard (applicable to full build)
A short intake, one 90-minute conversation, and within two days a written map: where your system leaks value, the three highest-leverage changes, and — only if warranted — a scoped proposal for a full build. Yours either way. This is the front door to a build — and the only door to one.
02 · The Extended Mind Buildout $1,500 founding rate / $3,500 standard
The complete bespoke system: work-life inventory, work-flow map, model constitution, a plain-text archive you own, valuable workflows turned into AI skills — some that draft in your own voice — and standing rituals that run themselves on whatever cadence fits your work, from a morning brief to a weekly reckoning, and a user's field guide. Your system runs in Orphic_OE — free, open source, local-first, and yours to keep whether or not I stay involved. What you're paying me for is the fitting: an egg written for how you actually work, and furniture built for your own workflows rather than a template with your name on it. Delivered in one to two weeks. 50% deposit, two revision rounds, finished when you can run it without me. 30 days of full stewardship included in build price. Founding price holds for the first three clients.
03a · Full Stewardship $350/mo
Unlimited break-fixes: deprecated models, skills that stop firing, integrations that silently break—it breaks, I fix it. Reply within 24 hours, fix aimed for 48 — a full day ahead of Lite at both ends. 4 hours a month I spend implementing your requested changes and updates. One structured session a month: review what worked, prune what died, adapt to whatever changed, and plan future developments for your system. The white-glove treatment—and you'll feel it.
03b · Stewardship Lite $250/mo
Just like with full stewardship, unlimited break-fixes when the AI landscape's weather causes damage—I'll respond within 48 hours and aim to implement a fix within 72, all told. 2 hours a month for requested changes and updates. A quarterly review session, where we look at what you're using and not using, and build towards the best system for you.
03c · DIY Stewardship $0/mo
Of course, if you choose not to have me tend your system after building it, it's still 100% yours, and will work just fine without me around—the system lives as plain text on your hard drive. A successful build will leave you with something that will remain useful long after my involvement.
Good fit: independent people with more context than their current system can hold, honest about the systems they've already tried and abandoned. Not a good fit: a fully autonomous agent to run your life, a productivity cure that changes no habits, or a substitute for therapy, coaching, legal, or medical care — I build working systems, and I stay in that lane.
What I've built against
Not a list of supported integrations — a list of formats and systems I already know how to read, write, and build furniture around. If you live in one of these, I've been there before.
- Notion
- Including the API — pages, databases, and the relations between them, read and written programmatically.
- Obsidian
- Databases built the fully Obsidian-compatible way, by design — Orphic_OE's own resolver follows the same rules on purpose.
- Plain files
- Markdown, text, and CSV in a filetree and wiki — the format every other integration on this list eventually gets exported down to.
- Calendars
- Google Calendar and Notion Calendar, read and written where the API allows it.
- Google Drive & Docs
- The wider Google ecosystem, named specifically rather than waved at.
Trust
Every prospect asks the same questions of any AI tool, and they deserve a straight answer instead of a privacy policy. Your material stays as plain files, not rows in a database I operate. I don't resell it or monitor it. At handoff you get everything, and I keep one archived copy myself — plain files on my own drive, not hosted access — gone if you ask. There's no server for a license agreement to quietly claim a stake in, because there's no server.
And there are things I won't build, on principle, not because the technology can't: no screen capture, no ambient audio, no browser-history scraping. A system that watches everything you do to save you the trouble of telling it what matters has stopped extending your mind and started replacing your attention. If that's the tool you're looking for, I'm not the one building it.
Where this stands
Modelwright Studio is early — the craft is real and already in practice with a first client's system delivered and another on the way. Posts on modelwrighting and MONET, co-authored with models, live on the aiBLOG.
If the idea is one you recognize the need for, the application's above. Questions first? Say hello.
Get in touch: xmh.dev@gmail.com