Free · living guide · for any discipline
The essentials of AI, to build a venture with it.
If you’ve never used an AI, this is the minimum so you don’t get lost —and if you already use it, the framework to use it with judgment—. Six ideas that don’t expire. It’s the same for a veterinarian, a biologist, or a psychologist: that’s why it lives here, in the series, and not inside a single book.
Generative AI is a model that learned, from enormous amounts of text and images, to predict “the most likely thing to come next.” It doesn’t “think” or “know the truth”: it produces what’s most plausible. Understanding this explains almost everything else —why it sometimes nails things astonishingly well, and why it sometimes invents with total confidence—.
A “prompt” is an instruction. The practical rule: talk to it like a brilliant intern who is literal and has no context. Give it the role, the context, an example of what you want, and the format of the answer. The clearer the assignment, the better the result.
Search engine: you ask, it answers, and it stops. Copilot: it works alongside you while you hold the wheel (it drafts, summarizes, corrects). Agent: it receives a goal and executes several steps on its own —it’s the leap that makes the “one-person company” possible—.
It invents with confidence (they call it “hallucinating”): it can hand you a quote, a fact, or a figure that sounds perfect and is false. That’s why the rule never changes: AI executes, you validate. No output touches a client, a patient, or an animal without passing through your professional judgment.
What you type can be used to train the model or get stored, and that changes depending on the version. In the free ones, your text can usually feed the training (sometimes you can turn it off in settings); in the paid, enterprise, or API ones, the norm is that it isn’t used for training and is deleted soon —but check it in each tool’s terms—. Don’t upload what you can’t afford to expose; aggregate or anonymize when you can. And an increasingly accessible option: run local models (open-weights) on your own machine, where the sensitive data never leaves your hands.
There are closed models you rent via API (fast and powerful, but you depend on the provider) and “open-weights” models you can run yourself (more control and sovereignty over your data). Don’t marry just one: treat compute like any other input, with a plan B.
The golden rule
AI executes; you —the professional— curate and validate. That’s the line no one should cross. AI multiplies you; it doesn’t replace your judgment. Teaching it without teaching this limit would be irresponsible.
This is what doesn’t expire. What does change every season —which tool, which button— we keep up to date here, so this guide doesn’t age. Want to learn to apply it to your project? Join the Founders’ Circle.
We’re building the community where professionals from every discipline learn to build ventures —with AI as the lever— and form teams. Leave your email: we’ll let you know the moment we open, and you’ll get more resources to start right now.
No spam, truly. We only write when there's something worth your time —series news and resources—. Unsubscribe anytime.