Beginner
Learn by building, not just reading.
Trails 01–03. Hands-on AI literacy. Short stops, real skill. Progress stays on this device.
Signal of the day
Logged for today.
Prompt lab
Four choices. One sharp prompt. Drop it into Trail 01 when you’re ready.
1 · Pick a role
2 · Pick a goal
3 · Pick context
4 · Pick constraints
Trail stop 01
Prompting like a builder
A weak prompt dumps the work on the model and hopes for magic. A builder’s prompt gives role, goal, context, and constraints — so the model has something solid to build against.
The four pieces
- Role — who should the model act as? (teacher, editor, coach…)
- Goal — what exact output do you want?
- Context — audience, situation, background that matters
- Constraints — length, tone, format, what to avoid
Builder: “You are a patient tutor for a 14-year-old. Explain what a large language model is in under 120 words. Use one everyday analogy. Avoid jargon; if you must use a term, define it.”
Your turn
Write a prompt that asks an AI to help someone prepare for a first job interview in an AI-related field. Use as many of the four pieces as you can.
Privacy tip: Don’t paste passwords, API keys, private student data, or anything you wouldn’t put in an email. “Check my prompt” stays on this page when possible; “Deeper coach feedback” may send your text to our site coach endpoint. Privacy
Feedback
Coach rewrite
Further
- Prompt Engineering GuidePatterns for clearer, more reliable prompts.
Trail stop complete
Prompt craft, locked in. Deeper coach feedback is there if you want a sharper pass.
Next up: Trail 02 — What AI is (and isn’t) · Trail 03 — Truth under pressure
Want more trails when they drop? Leave your email:
Trail stop 02
What AI is (and isn’t)
Modern chat tools are powerful pattern engines. They’re great at drafting, summarizing, and brainstorming — and they can also sound sure while being wrong. Builders stay useful by knowing the difference.
Hold these ideas
- Pattern, not person. These tools predict likely next words from training data. They don’t “know” you or live in the world the way people do.
- Fluent ≠ true. A smooth answer can still be incomplete, outdated, or made up. Confidence is not a proof.
- Temperature is a dial, not a mood. Many systems let you set “temperature”: lower values tend to stick to safer, more predictable phrasing; higher values explore more variety — and can wander into less reliable territory. Same model, different risk profile.
- You’re the judgment layer. AI can accelerate work; you decide what to trust, check, and ship.
Your turn — sort the claims
For each statement, choose Sounds right or Not quite. You’ll get a short explanation either way.
0 of 7 sorted
Further
- Elements of AIFree course for foundations without the hype cycle.
Trail stop complete
You practiced separating useful AI capabilities from common myths.
Next: Trail 01 — Prompting · Trail 03 — Truth under pressure
Want more trails when they drop? Leave your email:
Trail stop 03
Truth under pressure
People often push tools to agree, speed up, or sound certain. A strong response stays useful and honest — it doesn’t invent facts or surrender the truth just to please the user.
What “caving” looks like
- Agreeing with a false premise because the user insists.
- Inventing sources or numbers to sound complete.
- Overconfidence on high-stakes topics where caution is required.
- Holding the line means correcting gently, offering what’s true, and saying when something can’t be verified.
Your turn — pick the stronger move
Read each situation. Choose Hold the line (truth first) or Cave (please the user at the cost of accuracy). You’ll see why.
0 of 5 completed
Further
- Anthropic ResearchFor a serious read on model behavior and pressure.
Trail stop complete
You practiced choosing truth under social pressure — a core builder skill.
Next: Continue to Intermediate — Trails 04–06 →
Stay in the loop:
Earned Beginner activities are complete. This marks Trails 01–03. It is not a certification.
Trail Builders