Language Models: The Missing Manual
The mental model for what a language model actually is, starting from the blank box everyone already knows.
- Prompting
- Context
- When models fail
- Memory & limits
AI upskilling from first principles.
Not prompt recipes. The way the thing actually works.
I run hands-on sessions that build understanding from the ground up, from tokens to agents, so your team can reason about whatever AI throws at them next instead of chasing the last tip they read.
ScrollWhy first principles
Most AI training is a bag of tricks: prompts to copy, settings to toggle, tools to try this week. It works right up until the model, the interface, or the task changes, and then everyone is back to square one.
I teach the primitives instead. What a token is. What actually sits in the context window. Why a model forgets, hallucinates, or goes off the rails, and what an agent is really doing when it takes an action on your behalf. Once those click, the tips become obvious and the surprises stop being scary.
Understanding beats memorization. People who know how the thing works can figure out the tool they haven't seen yet.
How the sessions work
Sessions are live and interactive, built around demos you can poke at rather than slides you watch go by. We start from something everyone recognizes, the blank chat box, and build up one idea at a time until the internals feel familiar.
They work for a mixed room. Technical and non-technical people learn side by side, because first principles don't require a CS degree, just curiosity. I tune the depth and the examples to your team and the work you actually do.
I stay current with changes in the tools and across the industry, so you can make the most of what's possible at the leading edge from the first day we work together.
The sessions
A growing set of modules. They stand alone or stack into a progression, and each one can be shaped around your team.
From Tokens to Agents
Follow a single idea all the way up: how tokens become conversations, how conversations fill a context window, and how a chat quietly turns into an agent.
- Tokens
- Conversations
- The context window
- Intro to agents
Agents: Models That Act, Not Just Answer
What actually happens when a model does work on your behalf, so agents stop feeling like magic and start feeling like something you can debug.
- What is an agent?
- The agent loop
- Tool calls
- Context & compaction
Talking to Your Data
A language-model-first approach to data analysis with SQL, for people who know their questions better than they know their query syntax.
- Asking in plain language
- SQL you can read
- Trusting the answer
The same approach carries into one-on-one coaching. See examples of coding-agent adoption and long-term mentorship.
AI upskilling questions
- What is AI upskilling?
- AI upskilling helps people build the understanding and practical habits to use language models and agents effectively in their own work. The goal is durable judgment, not memorizing prompts for one tool.
- Who are the sessions for?
- The sessions work for technical, non-technical, and mixed teams. Examples and depth are adapted to the people in the room and the decisions or workflows they encounter.
- Do participants need coding experience?
- No. The first-principles material starts from familiar interfaces and builds one concept at a time. Coding-agent sessions can also be tailored to experienced engineers who want more advanced workflows.
- Can the training use our team’s real work?
- Yes. Hands-on sessions can be shaped around your team’s tools, domain, examples, and constraints so participants practice applying the concepts rather than only watching demonstrations.
- Is this a single workshop or a longer program?
- Either. Modules can stand alone for a focused workshop or stack into a sequence that moves from language-model fundamentals to agents, data, and role-specific workflows.
Want to bring first-principles AI education to your team? Let's talk.