💼 Business & applied AI
From a real pain point to a paying offer — and from prototype to evaluated agent. Families 08 (Product, Business, and Monetization) and 09 (Digital Product, AI, and Prototyping) apply the language to what generates revenue and what delivers it: market, customer, offer, price on one side; MVP, agent, eval, guardrails of the other.
By the end of the track, you’ll write a defensible business thesis and design the AI agent that executes it — with evaluation before scaling.
The track’s rule: the business thesis and the technical thesis go together—a product that charges but doesn’t work, or works but doesn’t charge, isn’t a product.
Course roadmap
💼 Product, business, and monetization
From /market to /growth: market, customer, positioning, offer, price, business model, and growth.
🤖 Prototyping and AI agents
From /prototype to /integrate: MVP, specification, agent, prompt, tools, memory, evaluation, and guardrails.
Detailed content
💼 Product, business, and monetization
The 10 commands in family 08 — from market analysis to growth loops.
Assesses a market's problem, demand, segments, competition, dynamics, and barriers—before any product idea.
Most products fail because there’s no market, not because there’s no code. Looking at the market first is the cheapest habit there is.
real demand · competition · dynamics · barriers to entry.
Define jobs (what the person is trying to accomplish), pain points, wants, context, current alternatives, and buying criteria.
You don't sell to a segment; you sell to a person with a job to do. Get the job wrong, and you build the wrong feature.
job to be done · current alternatives · purchase criteria.
/segment creates useful groups based on needs, behavior, or context; /position clarifies who it’s for, what problem it solves, which category it’s in, what makes it different, and the proof.
Positioning is the decision all others inherit — price, channel, messaging. Without it, every piece of marketing reinvents the product.
segment by need · category · difference · proof.
Organizes the promise, deliverables, price, proof, perceived risk, and call to action into a complete offer.
The product is what you make; the offer is what the customer buys. Good products with confusing offers don't sell.
promise · deliverables · proof · risk reversal · CTA.
Compares value-capture models, willingness to pay, unit economics, and sales simplicity.
Price is a business's highest-leverage decision and the one that gets the least methodical attention. A 10% change in price affects profit more than a 10% change in any cost.
willingness to pay · unit economics · pricing model.
Converts existing assets—audience, data, software, knowledge, or distribution—into revenue models.
Almost every business is sitting on assets that cost nothing. Monetizing means looking at what already exists and asking, “Who would pay for this, and in what format?”
latent asset · revenue format · billing channel.
/businessmodel maps value creation, delivery, and capture, costs, channels, and advantages; /moat assesses what prevents copying—brand, data, network, distribution, switching costs, scale, know-how, speed.
A business model without a moat is public proof that money can be made there—for anyone with more capital than you.
value creation/delivery/capture · moat · switching costs.
Identifies loops, channels, bottlenecks, metrics, experiments, and retention mechanisms — the system that grows the business without relying on linear effort.
Campaign-driven growth dies when the campaign ends. Loop-driven growth feeds on its own results.
growth loop · bottleneck · retention before acquisition.
🤖 Prototyping and AI agents
The 10 commands in family 09 — from the smallest testable artifact to an agent with evaluation and limits.
/prototype reduces the idea to the smallest artifact capable of validating the critical hypothesis; /MVP chooses the minimum set of capabilities that delivers value and generates real learning.
Every week you spend building before testing is a week betting that you guessed right. A prototype buys the answer at the lowest cost.
critical hypothesis · minimum artifact · learning before scaling.
Define requirements, interfaces, behavior, limits, exceptions, and acceptance criteria—the contract before the code.
The spec is the document that gets the model (or developer) to build what you intended, not what they imagined. Without a spec, every iteration is a surprise.
acceptance criteria · behavior in exceptional cases · out of scope.
Define an AI agent’s objective, tools, memory, rules, inputs, outputs, limits, and evaluation in a single design.
An agent without a project is a chatbot with tool access—the worst of both worlds. /agent forces you to make the eight decisions before the first line.
single objective · tools · memory · limits · evaluation.
/prompt specifies the role, objective, context, rules, tools, format, and quality criteria; /tools maps tasks to tools and defines when to use, not use, or escalate.
A prompt is the agent’s source code. A tool without usage rules is a button the agent presses whenever it wants—including when it shouldn’t.
role · rules · output format · when not to use the tool.
Separates session memory, profile, persistent facts, history, and update policy—what the agent remembers, for how long, and who can change it.
Poorly designed memory is an agent that forgets what matters and remembers what it should forget. An update policy keeps memory from turning into junk.
session vs. persistent · profile · update policy.
Define test cases, criteria, metrics, critical failures, and the passing threshold—the test suite that determines whether the agent can be rolled out to more people.
Without evals, every prompt change is a roll of the dice. With evals, it's an experiment with a result.
test cases · critical failure · approval threshold.
Creates rules for safety, authorization, validation, fallback, and escalation — what the agent doesn’t do, what it does when it doesn’t know, and when it calls a human.
Every agent will encounter a situation the project didn’t anticipate. Guardrails determine whether that becomes a "I don’t know, I’ll ask" or an incident.
authorization · input/output validation · fallback · escalation.
Maps systems, APIs, authentication, data, events, errors, and synchronization—how the agent communicates with the world that already exists.
A standalone agent is a demo. An integrated agent is a product—and integration is where 70% of the real effort lies.
API contract · authentication · error handling · synchronization.