Summary  
This chapter presents an iterative feedback loop—Ask, Understand, Produce, Verify—that guides you to craft effective prompts, interpret AI responses, implement code yourself, and validate your solutions for reliable learning and development.

General domain of usage  
Software development

## The problem with how most people start

Most learners open Claude and treat it like a search engine with better manners. They
ask a question, read the answer, feel informed, and move on. Two days later they
can't reproduce any of it.

The mistake isn't the tool. It's the mental model.

## What Claude actually does

Claude produces the most plausible next piece of text given everything it has been
shown. That single fact explains both its strengths and its failure modes.


**Plausibility** is the property Claude optimises for: how well a piece of text *fits* the surrounding
context, based on patterns across everything it was trained on. It correlates strongly with truth — which is why Claude is useful — but the two can come apart, and the gap between them is where every failure lives.


Definition


| Task | Why it works |
|---|---|
| Explaining a concept five different ways | Explanation is a language problem |
| Rephrasing something you didn't understand | It has seen a million rephrasings |
| Generating practice problems | Patterns of problems are highly regular |
| Reviewing your code or writing | Critique is pattern recognition |
| Unblocking you when you're stuck | It can always offer the next move |

**It is unreliable where "plausible" and "correct" come apart:**

| Task | Why it fails |
|---|---|
| Naming an exact library method | A plausible method name may not exist |
| Citing a specific page or paper | Plausible citations are easy to invent |
| Telling you the current version of anything | Its knowledge has a cutoff |
| Deciding what *you* should learn next | It doesn't know your goals |

Notice the pattern: it is strong on **process** and weak on **record**.


**Knowledge cutoff** is the date after which Claude has no reliable information. Anything about the *current*
state of the world — library versions, deprecations, pricing, recent releases, who
currently holds a role — sits past that line and is guesswork regardless of how confidently it's delivered.

The loop this course teaches

```
ASK  →  UNDERSTAND  →  PRODUCE  →  VERIFY
 ↑                                    │
 └────────────────────────────────────┘
```

- **Ask** — a well-formed prompt; **Understand** — at your actual level, with the analogy's limits;
- **Produce** — you write it, solve it, build it;
- **Verify** — you check it.

Claude makes **Ask** and **Understand** dramatically faster. It cannot do **Produce**
or **Verify** on your behalf. Every technique in this course is either accelerating the
first half or protecting the second half.


Master the skills needed to turn any lecture, PDF, or course into knowledge that
survives the week. You will build prompts that explain at your level, retrieval
loops that quiz you instead of lecturing you, flashcards that test recall, and a
verification habit that catches confident mistakes. Every technique runs on
material you already have to study, so by the end you have a weekly system
instead of a pile of chat history.

Task	Why it works
Explaining a concept five different ways	Explanation is a language problem
Rephrasing something you didn't understand	It has seen a million rephrasings
Generating practice problems	Patterns of problems are highly regular
Reviewing your code or writing	Critique is pattern recognition
Unblocking you when you're stuck	It can always offer the next move

Task	Why it fails
Naming an exact library method	A plausible method name may not exist
Citing a specific page or paper	Plausible citations are easy to invent
Telling you the current version of anything	Its knowledge has a cutoff
Deciding what you should learn next	It doesn't know your goals

Claude is a Tutor not an Oracle

The problem with how most people start

What Claude actually does