How AI Works
Build intuition for what's under the hood — how a model predicts, decides, learns, and where it goes wrong.
- 01Interactive lessonUp next
How an LLM Works
Type a sentence and watch the model predict the next token — then turn up the temperature and see it get creative.
~8 minNot started - 02Interactive lesson
Fair Decisions
Drag a decision threshold over real veterans and watch a fairness guardrail block an unfair model.
~8 minNot started - 03Interactive lesson
Precision & Recall
Move the threshold and watch the confusion matrix — every false positive and false negative is a real person.
~7 minNot started - 04Interactive lesson
Train a Classifier
Tune a model and watch its accuracy — and its fairness — change in real time.
~9 minNot started - 05Module check
Module 1 Check
A short quiz on how models predict, decide, and learn.
~5 minNot started
Prompting AI
The core skill of using AI well — write, test, and refine prompts to make a model do exactly what you need.
Building with AI
Go from a prompt to a working AI agent — composed from parts, wired together, and run.
Governing & Securing AI
Make AI reliable, honest, and hard to abuse — guardrails, grounding, and defense against attacks.
- 10Interactive lesson
How a Guardrail Works
Send a request through the input → model → output pipeline and watch each guardrail catch what it catches.
~7 minNot started - 11Interactive lesson
Catch the Hallucination
Check an AI's answer claim-by-claim against real sources, and see the citation guardrail flag a fabricated rule.
~7 minNot started - 12Interactive lesson
Defend Against Prompt Injection
Launch an attack, then toggle defenses and watch them block it — securing the AI system itself.
~8 minNot started - 13Hands-on lab
The Guardrail Console
Hands-on lab: trip each guardrail on purpose and watch every decision get logged.
~10 minNot started - 14Module check
Module 4 Check
A short quiz on guardrails, grounding, and AI security.
~5 minNot started
The Mission
Where it all ships — tools serve veterans and military families, and outcomes are tracked like a real program.
Learn to build AI you can trust.