Each track fixes one thing AI-built apps get wrong, and ends with a capstone spec for your own app.
- Tracks
- 7
- Modules
- 35
- Capstone specs
- 7
- Parts per module
- 10
Decision
The question the module answers.
Mission
What you will be able to build by the end.
Mental model
The main idea, in one paragraph.
Failure report
What went wrong, told as a short incident report.
Framework
The rules that would have prevented it.
Scenario
Choose what to build and see what happens.
Mistakes
The common mistakes, and why they happen.
Build task
Your spec: the plan for your own app, written with the module’s guidance.
With your coding agent
The prompt for Claude Code, Codex or Antigravity, and a checklist to check what it built.
Quiz gate
A short quiz. Pass it and the next module opens.
How your app is put together, how it saves data, and what happens when a step fails.
Why does an app that passes every test still break with real users?
A map of your app’s steps and what each one is allowed to change
How does an automated job go from request to checked result without skipping a step?
A six-step checklist for your automated jobs
How do you make sure an AI answer never saves wrong data or charges twice?
A check-before-save and no-double-charge plan
How do you save progress so a long job can pick up where it left off after a crash?
A save-and-continue plan for long jobs
System Verification Spec
CapstoneCan you bring the map, the steps, the checks and the save points together into one plan for your app?
Your app’s reliability spec
Set up multi-step jobs so they pick up from the last good point instead of starting over.
How do you lay out a multi-step job so you can always see which step ran and what it produced?
A step-by-step map of your most complex job
When a job splits into branches, how do you stop one branch from changing another’s data?
A plan for keeping branches apart
When a job fails halfway, how do you restart from the last good point instead of the beginning?
Save points and restart rules for your jobs
What happens to a job that keeps failing, without stopping everything else?
A holding area for failed jobs
Graph Engineering
CapstoneCan you design a multi-step job with clear steps, separate branches, save points and a holding area for failures?
Your complete job map
Rules that must always be true, checked wherever your app passes data from one part to another.
What must always be true in your app, and how do you keep it that way?
The rules for one connection in your app
When one part of your app fails, how do you stop the failure spreading to the rest?
A plan for keeping failures contained
When step 3 fails after steps 1 and 2 have already done something real, like taking a payment, how do you undo it cleanly?
An undo plan for multi-step actions
How do you change the shape of your data, or of an AI answer, without quietly breaking the parts that read it?
A plan for changing data shapes safely
Contract Engineering
CapstoneCan you write the rules for every point where your app passes data between parts, in a form that can be checked automatically?
Your complete data-rules spec
Retries and agent loops that always stop, slow down when a service is struggling, and set bad inputs aside.
How do you make sure a loop, like a retry or an agent, always stops, and notice when it is going in circles?
A stop plan for one loop
How do you retry a failing step without overloading the service it depends on?
A retry plan for one outside service
When a service you depend on is struggling, how do you stop your retries from making it worse?
A slow-down plan for a struggling service
How do you find one bad input that keeps failing a loop, and set it aside before it costs money?
A plan for setting bad inputs aside
Loop Engineering
CapstoneCan you design a loop that always stops, never makes a failure worse, and sets bad inputs aside safely?
Your complete loop spec
Bring the parts together into one app you can see inside of, and that recovers on its own.
How do data rules, job maps and loops come together in one app?
How the parts of your app connect
How do you keep a record of every decision your app and its agents make, so you can trace what happened?
A record-keeping plan
When something fails across several services, how do you find which step caused it?
A plan for seeing inside your app
How does your app recover from an outage, a bad database change or broken data without losing anything?
A recovery plan
Integrated Execution Systems
CapstoneCan you write the complete plan for an app that holds up with real users?
Your complete app spec
Work with Claude Code, Codex or Antigravity without losing control: what to tell it, what to allow, and when it must stop.
What belongs in your AI’s instruction file, and what should stay out, so the rules that matter get followed?
Your AI’s instruction file
How do you turn a decision into instructions your AI can build from, and check it built the right thing?
A build-ready spec for one feature
How do you let your AI work on its own without it running forever, spending too much, or making changes you cannot undo?
Rules for AI runs you are not watching
What should your AI be allowed to do on its own, and what must always wait for you?
Your AI’s permission list
Your Agent Workflow
CapstoneCan you write down how AI agents work in your project so a new agent, or a new teammate, follows the same rules?
Your AI workflow spec
The gaps AI-built apps most often ship with: leaked keys, users seeing other users’ data, AI features that take instructions from strangers, and fake packages.
How do you keep API keys out of your code, your app, your logs and your git history, and what do you do the day one leaks?
A list of your secrets and a plan for replacing them
How do you make sure every read and write checks that this user is allowed to see this record?
Access rules: who can see what
How do you build an AI feature that reads text from strangers without letting that text take control?
A risk map for your AI feature
How do you check that a package your AI suggests is real, is the one you meant, and is safe to install?
A rule for adding packages
Your Security Spec
CapstoneWhat must be true about your keys, access rules, AI features and packages before real users arrive?
Your security spec
Take one AI product from a validated problem to launch, passing a gate at every step. Included with lifetime access.
The Pain Gate
The Comprehension Ledger
The Infrastructure Inversion
The Observability Imperative
The 95/5 Inversion
The Changeability Trap
The Cost Boundary
Narrative Sovereignty
The Ship Gate
$29 one-time. All 35 modules, plus the ones we add.
14-day refund, as long as you haven’t completed more than one module.