Free account

Save your progress

Keep your skill report, learning progress and passport safe, and pick up on any device.

  • Report updates after every test
  • Learning progress on every device
  • Reserve your passport link
or with email
Free forever. No card. Your scores stay private unless you choose to share them.
Learn AI fundamentalsModule 1 of 4 · 0 done
  1. How AI models actually work
  2. Prompting that works first time
  3. Checking AI output
  4. Responsible AI at work
Module 1 of 4 · AI literacy · 25 min

How AI models actually work

Tokens, prediction, training and context windows — the mental model that explains both the magic and the mistakes.

By the end you’ll be able to

  • Explain next-token prediction to a colleague
  • Know what training data and context windows mean in practice
  • Predict which tasks AI is reliable at and which it isn’t

A very good next-word predictor

A large language model (LLM) reads text as tokens — chunks of words — and predicts the most likely next token, again and again. It learned those likelihoods from enormous amounts of text during training. Then it was tuned with human feedback to follow instructions and be helpful.

This is why it writes fluently about almost anything, and also why it can produce confident, fluent text that is false: plausibility is what it optimises, not truth.

  • Fluency is not accuracy
  • Models don’t look things up unless connected to search or documents
  • The same prompt can give different answers (sampling)

Context is everything it knows right now

Beyond its training, a model only knows what is in its context window: your prompt, the conversation so far and any files or search results attached. Give it the document and it can summarise it accurately; ask from memory and it may invent.

Where AI is strong and weak

Strong: drafting, rewriting, summarising given text, brainstorming, explaining, translating, classifying, writing code with review. Weak without tools: precise facts, recent events, arithmetic over many steps, citations, anything requiring accountability.

  • Use it as a fast first draft, not a final authority
  • Ground it in sources whenever facts matter
Practical exercise

Map AI onto your job

List the recurring tasks in a typical week of your work (or a job you want).

0/5 steps

Deliverable: A task map with three chosen AI candidates and quality criteria. Add it to your skills passport as a project when it’s done.

Checkpoint

Check your understanding

Answer 3 scenario questions. Get 2 or more right to complete the module.

1. Why can an LLM state false facts confidently?
2. What is in a model’s context window?
3. Which task is AI most reliable at without extra tools?
Measure it

See how your ai literacy scores

A free test adds AI literacy to your skill report and shows how close you are to the 75% job-ready target.

Go deeper

Free PDF · 74 pages · 2 editions

The Laws of UX, in one field guide

60 research-backed principles in 6 categories — what each one means, why it works on the human brain and how to use it on your own screens. Written by Sushant Singh, founder of Science of Apps, who has scaled consumer apps to lacs of users and spent 10+ years leading product across startups and digital businesses in the US, EU & India.

Pick your edition
Free PDF · 74 pages

Get Laws of UX by email

Choose your edition

You’ll also get occasional UX psychology tips. Unsubscribe anytime. Privacy policy

Free PDF · 90 pages · 2 editions

100+ Subscription Paywalls, decoded

110 real paywalls from top apps, taken apart, rated out of 5 and decoded with the Laws of UX — plus the frameworks, copy and checklists to ship your own. By Sushant Singh, founder of Science of Apps.

  • 110 paywalls rated
  • 34 Laws of UX
  • 12 frameworks

Which paywall scored 4.8 out of 5, and which scored 2.6? It’s all inside.

Pick your edition
Free PDF · 90 pages

Get 100+ Paywalls, Decoded by email

Choose your edition

You’ll also get occasional UX psychology tips. Unsubscribe anytime. Privacy policy