Agentic AI course

The agentic AI course where you build one real agent.

A good agentic AI course has you build an agent, not watch one. In AI PhD you build one in your own AI, for one errand from your real week, and every step is scored.

Free, about 5 minutes, no account needed. Not a university. Not accredited.

What an agentic AI course should teach

An agent is your AI doing one whole job for you, the same way every time, following rules you gave it. A course worth your time teaches the habits that make an agent trustworthy, and has you practice each one on real work:

• Answer from a source. The agent works from a document you give it, quotes the line it used, and writes NOT IN SOURCE instead of guessing.
• Use a tool well. You describe the one tool the job needs: what it takes, what it gives back, and what it does when it fails.
• Check its own work. You set the conditions a good run must meet, and the agent checks its plan against them before it acts.
• Do the real job. The agent runs once on real work, and you show the evidence: what it was given, each step it took, what it made and what its check found.

A course that only explains these leaves you with notes. One that has you do them leaves you with an agent.

What AI PhD has you build

One agent, for your errand: the one task from your own week you would hand an agent, the same task every week. Every step is scored the moment you paste back what your AI gave you.

1. The free entrance exam. Your AI turns a messy note from your week into a task list, and you pick your errand from it.
2. The classes. Each is one move, first on a sample and then on your errand, and what passes on your errand is saved as a part of your agent.
3. Your thesis. You propose the agent for your errand, then run it once on the real errand and show the run. Both parts are scored.
4. Your defense. An AI examiner asks about your own thesis and checks your answers, and a person at AI PhD reviews it. Pass, and you earn the AI PhD Certificate in Applied AI.

Open today

Every step that passed every check today.

1. Turn a messy note into a task list, then do it on your own week
2. Class 1: Make your AI stop guessing
3. Lab 1: The NOT IN SOURCE drill
4. Class 2: Give an agent a real tool
5. Class 1, part 2: Your errand's source rules
6. Class 2, part 2: Your agent's tool
7. Class 3, part 2: Your agent's self-check
8. Your thesis: an agent that does one real errand
9. Your thesis, part 2: run your agent on the real errand
10. Lab 3: The Three Item Audit

What it is not

Not a university. Not accredited. It is not a video course and needs no code. Each class is a ready prompt you run in your own AI, and what it gives back is scored here. What you build stays in your own AI.

What it costs

AI PhD Semester One is $97 once. No subscription, nothing renews.

A full refund inside 30 days, and you keep what you built in your own AI. A refund cancels the certificate, and its public record then says Revoked.

Start with the entrance exam. It is free, and it ends with your errand.