What's at Stake
The AI does the work; you carry the risk. Why this is the best time there's ever been to build — and why missteps cost more now that building got fast and cheap.
If you wanted to be an aerospace engineer, I'd tell you to buy an airplane kit and build it in your garage. The Playbook is that, for AI-native software — the pieces and parts laid out before you, optional build configurations to make it yours, and the hands-on teaching to make it stick. Edition 1 ships Q4 2026.
Lessons, exercises, working agents, configuration surfaces, telemetry, and dashboards — you move through the curriculum by working, not by reading chapters about it.
Judgement in two registers: the founder judgement that decides what deserves to be built, and the engineering judgement that decides whether it will hold up.
The Agent Development Life Cycle — the old software lifecycle, re-staffed by a crew of agents, with you holding the two jobs that never transfer. You start the work, and you answer for it.
Becoming viable is the short clock — get this idea into the world correctly. Remaining viable is the long one — stay the kind of builder who succeeds as conditions change.
The Playbook is a kit in exactly that sense — a private app you install and build through, a complete, worked example of an AI-native build with the full doctrine already applied inside it. The exercises ask you to produce real artifacts — a PRD, a pull request, a roadmap — and as you build, the app reads how you actually work, so you learn to ship with an AI-native posture and learn about yourself at the same time. Comprehensive enough to be used. Specific enough to be argued with.
It is grounded in industry best practices as they evolve, in my own day-to-day building as technical co-founder of Soaring Titan and Maximum Automotive Intelligence, and in the research stream you'll find on Remaining Viable. It will be updated as the field moves — because the field is not done moving.
A working outline — subject to revision as development continues. The four units are the four things you do with the kit. Each lesson pairs a position with a practice, a working agent or two, and an exercise that asks you to produce a real artifact. Not just what to believe, but what to do on Monday morning.
You install the kit on day one — before you feel ready — and it starts recording: every prompt, every accept, every reframe. Seven lessons on who does what: two kinds of intelligence at the table, and the two jobs that never transfer to the machine.
The AI does the work; you carry the risk. Why this is the best time there's ever been to build — and why missteps cost more now that building got fast and cheap.
What the machine is for, and what you're for. It decides how; you decide why — and only you can care whether the thing is actually right, not just whether it works.
Origination at the front, answerability at the back. You start the work and you answer for it; the agents take everything in between. The first question of the kit: did I choose this, or did it choose me?
Your condition is the binding constraint now. How to check yourself when you can't trust the mirror, and how to wire an alarm that fires on the days you'd rather ignore it.
The kit starts reading how you actually work — because the honest instrument is the one you could never keep on yourself.
The maintenance the new speed doesn't excuse you from — keeping the shop, the agents, and yourself in working order.
The first working session with the installed kit, and the habits that make everything after this unit land.
Unit 1 installed my operating beliefs as the lens. Unit 2 surfaces yours — which beliefs you chose, and which the vendors installed for you. The configured kit is the settlement: mapped to your situation, your stack, and the ambition that brought you here — and compiled into a shop you carry into every repo you own.
Growing a team, renting what you could own, depending on an outside shop, inheriting a mess you're scared to touch, shipping what you can't explain, or pushing a demo to a real product. Each drops a different job and makes a different trade.
The environment you already run — your stack, your tenure — surfaced as evidence about who you actually are, not how you'd grade yourself. And one permission worth granting: let the kit read what your own site already claims you're building toward.
A short, honest read that confirms your situation against what you've actually built, shipped, and inherited. A route, not a label.
The tool market, de-programmed: stable functional categories instead of vendor names — and which of your tool beliefs you chose, versus which were installed for you.
For every skill in front of you, one call: own it for real because you'll live with it, or borrow just enough to direct it and stay in charge. The trap on each side, and the good news.
The universals every configuration stands on, whatever your situation — the conditions no configuration gets to skip.
Every escape is a trade — no move sheds risk, it relocates it. Borrow across situations on purpose, and choose the dependency worth holding instead of waking up inside one.
The configured kit is the settlement — your beliefs, tested against your behavior and the installed doctrine; your aim, written in your words with your own public evidence laid beside it. The settlement compiles into a shop you take with you. And what to do Monday morning.
The configured kit, run against real work. The Agent Development Life Cycle — the software lifecycle you may never have been taught the name of, re-staffed by a crew of agents, with you on the two gates that never transfer.
The path every app on your phone was built on, and why each step existed. The path stays — the team in the middle becomes a crew of agents, producing everything a full human team once did.
One place the work lives. A set of specialized agents, each with one narrow job. The house rules that make them build it your way. The adapter that lets them see everything the company already knows. And the same shop, compiled to travel — it goes with you into repos the kit never sees.
Follow a single piece of work across the board, left to right. The agent fills the column; you own the gate between the columns. That rhythm is the whole job.
Where you start, and where you answer — and the in-flight move (steering) that keeps both gates real when the middle runs at machine pace. Because agents absorb the time the middle used to cost, real tests, granular security, and solid infrastructure become the floor — not a late-stage luxury. And the one place that claim breaks.
The agent will reach for the frontier model on every task unless you configured otherwise — and your token bill, your rate-limit exposure, and your provider dependency will climb without you ever signing for them. The classify-and-route discipline that meets that default, with open-source and self-hosted models as a first-class option.
The same rulebook can make you sharper or hollow you out. Set your agents to explain where the skill is yours to own; set them to just produce where you'll only ever direct. The knife-edge under every teaching agent — and the same setting that decides whether you can see into the middle while the work is running.
Because the whole build runs through one place, it quietly records how you worked — what you reframed, what you actually read before approving, what you chose versus rubber-stamped. The honest instrument you could never keep on yourself.
Which gate to stand on first, and how the rest of the kit backs you up while the rhythm becomes yours.
Becoming viable is this build. Remaining viable is every build after. Everything the kit has recorded since Unit 1 becomes a dashboard you read in years — and it arrives seeded with my own readings, part of what $495 buys, until your own data takes over.
Every reading on the short clock flatters you — a shipped feature, a happy client. The second clock ticks in years, and it's the one almost nobody is watching.
Trends, not snapshots — read against your own baseline. The most honest readings come straight from your work: your transcripts — and, once connected, your invoices and your last ten bios — not your gut on a given afternoon. Every asked question arms with granular consent.
Three families — the builder, the craft, the ground. Nine things actually under threat over a career, each one planted earlier in the kit and coming due here. A working translation of the framework from Remaining Viable.
What to actually do when a dimension reads unfavorable — strategies, not affirmations, for each of the nine.
The quarterly reading protocol: how to sit down with your own instrument, read trends against your own baseline, and resist the urge to grade yourself kindly.
A bad reading isn't a verdict; it's the next thing worth originating. The loop back to the first gate.
The cohort opt-in: extend the same readings across builders running the same kit, with granular consent. Never required — the kit works identically if you decline.
Remaining viable feeds straight back into becoming viable — and the whole loop starts over, one build smarter than last time.
Lessons and ordering will shift as drafting continues. The intent — comprehensive, prescriptive, honest — will not.
Install. Configure. Use. Watch.
Quarterly cadence — because the field doesn't move at annual speed. Each edition incorporates what's been learned since the last — dated, versioned, and visible, with a changelog you can read. The configuration you build in Unit 2 survives minor revisions; breaking changes carry a migration the kit performs, not you — and the portable shop it compiles simply recompiles from your settlement, never hand-patched.
Buying once includes a revision window — the exact terms will be stated plainly when Edition 1 ships. No subscription, no upsell. The intent is a living body of work you can return to as conditions shift while you are using it.
Pre-orders open closer to release. In the meantime, leave an email and I'll let you know the day Edition 1 is available — and only that day. No drip campaign.
One-time. Private kit app, integrated lessons, exercises, working agents, telemetry, and dashboards — and the portable shop your configuration compiles for your own repos. Revision window included — terms stated plainly at release.
No charge today. I'll email once when Edition 1 is ready, and once when pre-orders open.
The Playbook is the self-directed artifact. Resources shows the doctrine applied to real, publicly-sourced problems. Mentorship applies the same kit to your problem, in the room, with me sitting next to the person doing the work. Each one is complete on its own. They reinforce each other for the leaders who want all three.
If you want the engagement behind the doctrine → Mentorship.
New Resources, Research, Data readings, and Playbook editions — this list is the only announcement I send.