What we teach, in the open
These simplified versions are the sessions we run with executive teams and champions inside companies, published as short click-through lessons. We think the whole world needs to understand AI better.
These lessons are designed to get you to a basic understanding of what’s needed to accelerate your skills and your team’s AI development. This is not all-encompassing, but it is an intentional grounding in the technology.
What is AI, actually?
Deterministic to probabilistic: the kind of computer that actually got invented, the trade it makes, and why the harness around the model, not the model itself, is what made AI useful.
The five kinds of data
Every answer an AI gives you draws on one of five sources: training data, your prompt, connected systems, the web, and memory. Knowing which one you are leaning on, and how far to trust it, is most of the game.
Understanding hallucinations
AI makes mistakes. Why hallucinations happen, the size effect that ties the error rate to the token count of a step, and why reasoning reduces errors without eliminating them.
Anatomy of a simple system
Frontend, backend, datastore: the three parts nearly every system is made of, the endpoints that connect them, and one real request end to end, a call to an LLM to proofread text.
Connection types: API, MCP and CLI
How systems talk to each other and how AI joins the conversation: APIs built for the deterministic era, MCP as the AI-friendly wrapper that bundles discovery, instructions, and authentication, and the CLI as the local toolset agents can run.
Understanding basic file types
CSV, JSON, MD and TXT: the four plain-text formats that carry most of the data you will hand to an AI, what structure each one adds, when each one fits, and why to avoid heavyweight formats as input.
Understanding auth
How systems know who is asking and what they are allowed to do: username and password, API keys, magic links, and OAuth, plus 2FA and the rules that keep credentials safe when AI is doing the work.
Relational DB basics
Tables, unique IDs, and relations: how structured data is really organized, why each fact is stored once, and enough SQL to see what any system is doing.
Understanding servers
What a server actually is, what the cloud really rents you, why the database you pay for is one you will never see, and the self-hosted other direction: the real choice is who operates it.
Vectors and RAG
How a machine finds text by meaning instead of keywords: embeddings that turn meaning into position, vector databases that find the nearest neighbors, and RAG as a retrieval flow in which vector search is one tool among several.
Loops, panels and error checking
Every model gets things wrong. The two kinds of wrong, the free check a parser can do, the retry loop and its cap, the panel of cheap models that votes on the judgments code cannot make, and the answers that cannot be checked at all.
Understanding Git
Version control keeps every version of a project and lets you return to any of them: commits, repositories and branches, why Git is not GitHub, how code is published and released from a repository, and why it is what makes letting an AI change your code safe.
Understanding packages
Almost no software is written from scratch. What a package manager installs on your behalf, why ten packages become several hundred, what a lockfile is for, the questions worth asking before adding one, and what to check when an AI adds them for you.
A reference architecture
Coming soonArchitecture first
Coming soonThe shapes a tool can take
Coming soon+ 9 more lessons in the queue
These lessons are excerpts. The full sessions run inside your company.
We teach these to your executive team and your champions, on your data and your problems, then stay to build. That's the in-company acceleration program.