Taught to you
Run a small AI team that ships real software in your domain.
The playbook a finance professional used to ship fifteen production systems in eight weeks. Without writing the code.
The problem nobody talks about.
Most AI training in the market teaches you how to prompt a chatbot. The ceiling on that is low and shrinking as models get better.
The unmet need is one layer up: how to operate an AI team that ships durable work over weeks and months, with memory, with reversibility, with cost control, without you writing code yourself.
That's what this is for.
The proof point
What this framework produced.
I'm a finance professional. Investment analyst by day, fractional CFO and side-project builder by night. Self-described weak coder.
In eight weeks I shipped roughly fifteen production systems:
- A multi-agent prediction-market analysis engine running on a fleet of LLM agents
- A financial data portal with Monte Carlo simulation, DCF valuation, and options analysis
- A trading bot family with full paper-first discipline and reversibility
- An ML-based decision engine using gradient-boosted trees on engineered features
- A multi-agent gateway connecting tools through a Telegram interface
- A watchdog with PID locking covering every running process
I wrote none of the code. I wrote the specs. I made the decisions. I approved the rolls. The agents did the work.
The unlock was not better prompts. It was a three-agent team with shared memory, routing rules, and an ops doctrine. That's the framework I'm sharing.
What you'll learn
The framework in eight pieces.
The Three-Agent Stack
Why three agents (not one, not ten), and how to assign them across tools you already pay for.
Routing by cost class
The single biggest cost-cutter. API tokens for thinking, subscriptions for grinding, free tier for relay.
The Shared Vault
An Obsidian folder becomes a cross-agent brain. Memory survives across sessions, agents, and months.
The Memory Lattice
Four memory types and how to structure them so they compound instead of rot.
The Ops Doctrine
Seven steps every change runs through. Catches AI mistakes before they hit production.
Autonomous Iteration
The Codex /goal pattern done right. Behavior contracts, non-goals, verification paths.
Research Swarms
When the question is hard, spawn 14 agents. The synthesis beats any single agent answered 14 times.
Watchdogs and Heartbeats
Process supervision, PID locking, scheduled checks. The infrastructure that lets you sleep.
What's in the package
Nothing missing.
Pricing
Three tiers.
Pick the level of help that matches how fast you want to be operational.
Playbook
$397
Self-paced. The whole framework, ready to read and fork.
- 14-chapter playbook
- Vault template
- All memory and process templates
- Case study and demo walkthrough
- 12 months of updates
- Email support, batched
Playbook + Onboarding
$1,997
Live setup help to skip the week-two friction.
- Everything in Playbook
- 90-minute onboarding call
- Pick your three agents together
- Build your routing tree live
- Seed your first feedback memories
- Async vault review afterward
- One week of email Q&A
Done-with-You
$12,000
4 to 6 week engagement. We do it together.
- Everything above
- Weekly working sessions
- First proof project built jointly
- Watchdog and notifications wired up
- Domain-specific playbook addendum
- Limited to 2 to 3 per quarter
FAQ
Common questions.
Will this work in my domain?
Probably, if your work involves recurring tasks that benefit from memory and reversibility. Examples that work: fractional CFO work, accounting automation, agency operations, content production, real estate analysis, paralegal workflows, recruiting. Examples that don't: one-shot tasks with no continuity.
How much will it cost to operate?
Subscriptions you'd pay anyway plus $10 to $50 of API per month. Most operators run the whole stack for $50 to $100 monthly.
Do I need to code?
No. The whole point. You drive the team. The team writes the code.
How does this relate to AI Setup?
AI Setup is "build it for me." The Operator's Stack is "teach me to operate my own AI team." Some buyers do both. Some start with one and add the other later.
Refund policy?
14-day no-questions-asked refund on the Playbook tier.