Stop pasting your business into the chatbot. Build a system that knows it.
A four-weekend intensive for owner-operators who are tired of re-explaining their business to a chatbot that forgets everything by Tuesday. You leave with a working AI OS on your machine, the workflows that recover your week, and the maintenance habits that keep it useful for years.
Format
Live cohort · 4 weekends
Built for
Owner-operators
Outcome
A working AI OS
Status
Waitlist open
// the decisions running your week
who gets answered today, who waits, who never should have made it to your inbox. Made by mood, by the order Gmail loaded, by whoever shouted last.
Source: The decision behind the inbox
which inbound leads are worth a real reply, a fast no, or a careful follow-up. Currently made in the seven seconds you spend on a name you don't recognise.
Source: The decision behind the pipeline
what you already know about this client, this deal, this draft. Currently re-derived from scratch every Monday because nothing remembers it but you.
Source: The decision behind the work
You don't have an AI problem. You have a memory problem.
Every ChatGPT or Claude conversation starts from zero. You re-explain your clients, your tone, your offer, your last meeting, for the fifteenth time this week. The output is generic because the input had to be.
So you copy-paste it into Gmail, Notion, your CRM, and call it productivity. It isn't. It's a more articulate version of the same admin treadmill.
From workflows you run, to decisions the system tightens for you.
Workflow thinking
Same work, faster
- 01Map every process, score impact and effort
- 02Automate the top quartile, declare a 38% time saving
- 03A year later, the company is unchanged
- 04Decisions still made by mood and accretion
- 05Generic outputs to specific problems
Decision thinking
Different work, made consistently
- 01Pick the four decisions that shape your week
- 02Tighten each one against your real context
- 03The system reads your wikis, drafts, history
- 04Hard rules enforced by hooks, not hope
- 05Hours come back as a side effect, not the goal
Four weekends. Eight live sessions. One system that runs.
01Foundations: the vault, the manifest, the first decision
By Sunday night, the most repetitive call in your week is being made the same way every time, against your real context, before you sit down.
Saturday · 2 hrs
Why current AI workflows fail, and what replaces them
- ·The three structural failures of chatbot-style AI
- ·Live demo: an email becomes a drafted reply, an updated record, and a logged decision in 90 seconds
- ·The three-layer memory model (raw, wiki, outputs)
- ·You set up your vault, on your machine, with your real files
Sunday · 2 hrs
Your manifest, and your first working command
- ·Write CLAUDE.md, the runbook every session reads first
- ·Your first slash command: /triage-inbox
- ·Run it live on your real inbox
- ·Three peers demo theirs; we debug together
Takeaway
A fully structured vault, your personal AI manifest, and a working morning triage running on your real email, by 7pm Sunday.
02Decisions: the four choices that shape your week
The four recurring decisions where owner-operators leak the most value: who to qualify, who to follow up with, what to recommend, what to record. You tighten two of them against your real business, running on Monday.
Saturday · 2 hrs
The qualification decision · the follow-up decision
- ·Which inbounds are worth a real reply, scored against your real history, not a generic rubric
- ·Which conversations get a follow-up and what that follow-up should say, drafted from the meeting note your system already wrote
- ·Variants demoed for realtors, advisors, agencies, coaches
- ·You implement both against a real (or simulated) client of your own
Sunday · 2 hrs
The recommendation decision · the recall decision
- ·What to propose to this client, scored against your past wins and your stated voice
- ·What you already know about this deal, surfaced before you re-derive it from your inbox
- ·You pick the two decisions where tightening pays back fastest in your business
- ·Cohort sanity-checks each other's choices against the loose-and-tight test
Takeaway
Two of your week's load-bearing decisions made the same way every time, against your real context. Hours come back, but the durable change is consistency.
03Specialists & integrations: connect what you already use
The system stops being a clever toy. Your decisions are now made against live signal, in your real tools, in the moment they happen.
Saturday · 2 hrs
Build your five-agent team
- ·Researcher (web access, no shell)
- ·Writer (no web, can't hallucinate sources)
- ·Editor (read-only on source, write-only on drafts)
- ·Librarian (read-only, the vault's memory)
- ·Optional bookkeeper (financial files, never anything else)
Sunday · 2 hrs
Connect Gmail, Calendar, Drive · ship one custom skill
- ·Authenticate the three core MCPs
- ·Upgrade your existing workflows to draft directly in Gmail
- ·Build one packaged Skill specific to your business
- ·Templates for service, e-comm, professional services, creator
Takeaway
A specialist agent team and live integrations. Drafts now appear in Gmail, not in chat windows you forget to check.
04Hardening & the long run: make it bulletproof, make it last
The difference between a cool demo and a system you trust with the decisions it now makes for you. Guardrails, scheduled runs, and the maintenance habits that compound for years.
Saturday · 2 hrs
The 70/100 problem · guardrails that don't depend on hope
- ·Why CLAUDE.md rules are about 70% reliable and hooks are 100%
- ·Build four essential hooks: immutability, confirmation, summary, cost
- ·Industry guardrails: redaction for lawyers, financial-data rules for accountants
- ·Workshop your own industry-specific guardrail
Sunday · 2 hrs
Scheduled runs · maintenance · demo day
- ·Two scheduled automations: morning draft, weekly review
- ·The 20-min/day, 1-hr/week maintenance contract
- ·Your written 90-day rollout plan
- ·Demo day: every student presents their best workflow
Takeaway
A system with safety rails, scheduled drafts ready for your morning review, and a written plan for how this grows over the next year.
“The owners who shape their context layer nowwill look back on this moment the way early operators look back on the first cloud tools.”
— The premise of Cohort 01
Not theory. Not a notion of what you could build. A working system.
→ 01
A private AI OS on your machine
Tailored to your business, your voice, your clients. Yours forever, regardless of what model you point it at next year.
→ 02
8 to 12 custom commands
For your real daily operations: triage, research, drafting, follow-ups, weekly review. Built around how you actually work.
→ 03
Five specialist subagents
Researcher, writer, editor, librarian, bookkeeper. Each with their own tool boundaries, so they can't do what they shouldn't.
→ 04
Live integrations with what you use
Gmail, Calendar, Drive, plus a packaged Skill specific to your industry. Your AI works inside your real tools, not parallel to them.
→ 05
Industry-specific cheatsheets
Realtor, advisor, agency, coach, e-comm, professional services. CLAUDE.md additions and workflow mappings for your category.
→ 06
A written 90-day plan
How this grows from the cohort into your operating reality. Daily habits, weekly cadence, quarterly review checkpoints.
Join the waitlist. 12 seats.
Cohort 01 is small, deliberately. We open applications to the waitlist first, owner-operators serious about building this, not the AI-curious.
One email when applications open. Date, price, application link. No other emails.
Cohort
01
Seats
12
Format
Live
// waitlist
Hear first when Cohort 01 opens.
Waitlist gets first read on the twelve seats. Date, price, and the application link land in your inbox the day they go live.