Courses
Practical courses to help you build better with AI.
Courses
Beginner
Start here: core ideas, vocabulary, and first useful moves.
Personal Assistant
Install the apps, set your preferences, connect your accounts, and save repeatable skills.
Installing Claude and Codex
You will get Claude and Codex onto your computer, signed in and up to date, and you will know how to judge which one fits a task on your own work, since their strengths keep changing.
Getting Your Settings in Order
You will turn off data training, turn on memory, pick a sensible model, set how much the assistant can do without asking, turn on speech, and save one instruction that makes it teach you as you go.
Connecting Gmail, Calendar, and Transcripts
You will connect the three accounts that matter most the safe way, by just asking. You will avoid the connector mistake that half-works and keep email to draft-only.
Drafting Emails in Your Voice
You will learn what a skill is, capture your writing voice from your own emails, and build a reusable skill that drafts a follow-up in your voice as text you copy. It drafts; it never sends.
Pulling Enormous Value From Your Meeting Transcripts
You will build a reusable skill that reads a meeting transcript and pulls out the summary, the key takeaways, the decisions, and the next steps, each one carrying the lines it came from, then goes further: it organizes the talk into categorized tasks, coaches you on what the text shows about how you communicated, and lays out options for solving the problems the meeting raised.
Turning Results Into a Beautiful Page
You will ask for one page you can open with a double-click, direct its look in plain words until you love it, the fonts, the colors, the spacing, and the type, add a diagram or a small interactive piece when it helps, and save the whole routine as a reusable skill.
My First Skill
You ran a summary on your own notes, saved the working process as a named skill, and ran it on a fresh set of notes to get the same three labeled lines back.
My First Meta-Skill (A Skill Router)
You built a router skill, write-in-my-voice, that chooses between separate voice skills based on the request in front of it.
Connectors
Connect your assistant to a real account and prove it can read the source.
Prompting
Request design, examples, context, and better agent direction.
Version Control
Git, GitHub, commits, rollback, and safe shipping habits.
Intermediate
Build on the basics with multi-step workflows and quality checks.
Clean Agentic Building
Rules, standards, source-of-truth systems, and drift control.
- Members only
Single Source of Truth for AI Tools
You will be able to keep one home for your AI rules, split it into always-read and by-project files, point every tool at it, and catch stray copies before they pull your tools apart.
Connectors and Plugins
Reach services with a ready-made connector or plugin, and the ones that need their own key.
- Members only
Connecting When There's No Connector
You will connect a service that has no official connector by using its own API key: you will get the key from the service itself, give it only the access the job needs, keep it secret, test it, and revoke it when it leaks or you are done.
Dynamic Workflows and Subagents
Direct teams of subagents and build multi-phase workflows on the fly.
- Members only
Introduction to Dynamic Workflows
You will be able to describe a multi-stage job to your agent and have it assemble the work for you: stages that run in order or side by side, helper subagents that take the parts that fan out, a fresh helper that checks the work, and a final step that combines everything into one result.
My First Dynamic Workflow with Subagents
You asked one lead agent to split a job into independent parts, run those parts with subagents in parallel, and return one result per part.
- Members only
Quality Ratcheting With the Tournament Workflow
You will be able to run a small contest for any open-ended piece of work: ask the agent for several versions in different directions, judge them against what the work has to do, then combine the best parts into one final version that can come out stronger than any single try when your judging list is solid.
- Members only
Verifying Work With an Independent Verifier
You will be able to set up an independent verifier: a fresh subagent that did not make the work, handed your standards and told to find where it falls short. You will give verifiers different angles, trust the problems they agree on, and loop verify-and-fix until the work meets every standard.
My First Subagent Verification in the Loop
You had an independent verifier subagent check a finished piece of work against your standards, fix what it found, and hand back one corrected result.