How much AI fluency do you have?

AI fluency isn't knowing how to use ChatGPT. It's understanding how AI works under the hood well enough to make good decisions when you build with it: which model to pick, why it hallucinates, when an agent is overkill and when a simple prompt is enough.

For a product engineer it's the difference between asking the AI and praying, or directing it with judgment. The first leaves you at the mercy of results you can't evaluate. The second turns you into someone who multiplies their work instead of delegating blindly.

The questions span five areas, one from each on every attempt:

• What an LLM is and how it thinks — prediction, hallucinations, training cutoff.
• Tokens, context and cost — why you pay what you pay and what temperature controls.
• Prompting — system prompts, specificity and iteration.
• Agents, tools and autonomy — what separates an agent from a chatbot.
• Integrating AI into your product — workflow vs agent, RAG, human in the loop, observability.

Almost every expensive mistake comes from the same misunderstandings:

• Trusting an answer because it sounds confident — a model hallucinates with the same poise it gets things right.
• Building a RAG or a multi-agent system when a prompt with context would have been enough.
• Measuring nothing: when something breaks in production, there's no way to know what happened.

Recognizing these traps is half of AI fluency.

Knowing what a token is is good. Knowing how to turn that into a product that charges money is another level. That jump — from understanding AI to building with it — is the heart of the Founding Product Engineer training.

If this test feels too easy, you're probably ready for what comes next.

This test's questions come from 100 Questions to Understand LLMs and Agents. If you miss one, the full explanation is there. And if you want to go from theory to a real product, the courses take you by the hand from idea to deploy.