From a probabilistic system
Can I cancel my flight and get a refund?
answers vary each run
The same question, and the answer changes each run.
AI-Native Product Management
Learn to build and shipAI products you can trust.
Build AI-native products with confidence, from the frameworks behind great decisions to the hands-on skills needed to ship production-ready work.
To a product you can trust
Can I cancel my flight and get a refund?
Your Basic Economy fare is not refundable after 24 hours, but you booked six hours ago, so you can cancel now for a full refund. Do you want me to cancel it?
- grounded in your booking
- checked against fare rules
- evaluated before release
The same question, one answer, and a way to show why it is right.
The Builder's Stack
The AI course builders wish existed. Pick your level and go.
Level 1The Fundamentals~8 hrsLearn how software works, then direct an AI build of your own from idea to first real users.The Software MapThe Technology StackThe Model LayerShip and Track
Level 2The Practice~8 hrsMake a working build safe, tested, and ready for real users.Experience Design Principles for AI ProductsEvalsAgentic systemsSecurity
Level 3The Frontier~12 hrsTake on the newest AI problems, from agent fleets to agent users.Multi-Agent OrchestrationAI in high-stakes industriesContext & MemoryThe Unit Economics of AIWhen Your Users Are AgentsRunning an AI Organization
Or enter the stack from the top →see it applied in the FuelTheFam case study →
The Framework
The cycle that turns a probabilistic model into a reliable product.
Shape · Ship · Track is our delivery cycle for AI products. Each step below shows the work inside it.
Continuous Operations
- Govern the knowledge
- Govern access and safety
- Supervise the agents
- Build the team
A probabilistic model
Shape
Decide how it behaves
Frame the problem
Write the behavior
Choose the model
Prototype it
Ship
Put it behind guardrails
Build the guardrails
Build the eval suite
Earn trust when unsure
Set the cost budget
Track
Catch what users won't report
Watch it in production
Catch the drift
Feed it back into Shape
A reliable product
Field Notes
Lessons from the frontier of building AI products.
September 3, 2026 · 6 min
How We Budget an Agent Fan-Out Before We Run It
After we let a Fable 5 loop use up most of a week's allowance, every fleet we launch now gets an estimate built from a measured unit, a hard cap set in the same message as the approval, and a pilot before the rest runs. On our largest run so far the unit held and the agent count is what the estimate missed. This is the routine, and what we changed at each miss.
September 2, 2026 · 6 min
Fable 5.1, First Impressions: Less Steering, Faster Spending
We gave Fable 5.1 a seven-chapter rewrite on its first day, with Claude Code's fleet mode on. This note covers what got better, why its estimate of its own run was short by a fifth, how a five-hour allowance ran out in 43 minutes, and how far the writing has come.
August 16, 2026 · 5 min
Why We Stopped Letting AI Pick the Winner
For a few weeks we let Claude generate creative options and let a judge agent pick the best one. The picks were competent, and we rejected nearly every one. We also cover what the judge still does for us, and why we took the final pick back ourselves.