At the end of his 47-minute lecture, Geoffrey Hinton said this:

> “If you sleep well tonight, you may not have understood this lecture.”

He built the neural networks that underpin all modern AI, then left Google to publicly warn that AI already outperforms humans on most cognitive tasks.

Yet most people open Claude, type one sentence, close the tab, and think they're using AI.

In reality, they're only using roughly 10% of its capabilities.

Anatoli Kopadze distilled Hinton's talk into 17 Claude features that 99% of users have never discovered. Below we highlight the most useful ones you can start using right away.

---

### 1. Projects — Claude Finally Remembers You

Every new chat starts from scratch with Claude. It doesn't know your name, your job, or your preferences. Most people accept this and reintroduce themselves every single time.

Projects solves this problem. Create a project, upload your documents, and write permanent instructions. Claude will remember everything. When you open it next week, it picks up right where you left off.


Paste the following into your project instructions, and it will work immediately:

```

You are my content research assistant. I run a newsletter about AI and crypto for a technical audience.

Always assume my readers know the basics. Don't explain what an LLM is or what a blockchain is.

When I share a topic or article, your job is to:

1. Identify the 3 most counterintuitive or surprising angles

2. Find connections to recent events I might have missed

3. Suggest how I could frame this as a story, not a summary

Tone: direct, no corporate language, no filler phrases.

Format: short paragraphs, no bullet points unless I ask.

Never start a response with "Great question" or "Certainly".

```

### 2. Artifacts — Run a Working App Right Inside Your Chat

Many people think Claude can only output text. That's wrong. Artifacts lets Claude build fully functional tools in the sidebar — calculators, habit trackers, games, dashboards. It supports SVG, interactive charts, and Mermaid diagrams. It's even available on the free plan.

Try this prompt:

```

Build me a habit tracker as a working web app.

I want to track 5 daily habits.

Each day I can check them off.

Show a 7-day streak counter for each habit.

If I miss a day, the streak resets.

Design: dark background, clean minimal look.

Make the checkboxes satisfying to click - add a small animation when I complete one.

The data should persist if I refresh the page.

```

### 3. Adaptive Thinking — Make Claude Actually Think

Most people have never flipped this switch on. In Extended Thinking mode, Claude reasons step-by-step, and you can see the entire process.

You don't need it for simple questions. But for complex decisions and strategic analysis, the difference in output quality is night and day.


Use this when you're facing a real decision:

```

I'm deciding between two options and I want you to think through this carefully before answering.

Option A: [describe option A]

Option B: [describe option B]

My situation: [your context, constraints, what matters most]

Work through this before responding. Think about:

- The second and third-order consequences of each option

- What I'm probably overweighting or underweighting emotionally

- What information I might be missing that would change the decision

- Which option has better downside protection if things go wrong

Then give me your actual recommendation with your reasoning.

```

### 4. Memory — Never Introduce Yourself Again

Turn on Memory, and Claude will build a profile of you over time: your job, your projects, your communication style. It already knows who you are when you start a new chat.

It's off by default, and most people don't even know it exists.


```

I want you to remember the following about me so you don't need to ask again:

My name is [name]. I work as [role] at [company or project].

My main focus right now is [what you're working on].

My audience or customers are [who they are].

When I ask for help, always assume this context unless I say otherwise.

My preferred communication style: [direct / detailed / casual / formal].

Things I find annoying in responses: [e.g. bullet points, long intros, excessive caveats].

Save all of this to memory now.

```

### 5-9. Role Prompting — One Prompt Changes Everything

Claude doesn't have to be a generic "AI assistant". Give it a specific role, and it completely changes how it answers questions, pushes back, and what it tolerates.

**Therapist**: Instead of giving direct advice, helps you work through cognitive distortions by asking questions.

**Tough Mentor**: Stops agreeing with you and stress-tests your plan.

**Personal Trainer**: Builds a 12-week plan based on your actual data.

**Conversation Practice**: Plays the role of a difficult person so you can practice before a real conversation.

**Devil's Advocate**: Builds the strongest possible case against your already-decided position.

---

### 10. Claude in Chrome — It Sees What You See

It's a browser extension. Claude can read your current page, click links, fill forms, and flip pages. Just describe the task in natural language and leave it to work.

*Search for Claude for Chrome to install → log in → click the extension icon to open the sidebar.*

### 11. Claude Cowork — It Lives on Your Desktop

It's a desktop app that can access your file system directly. It reads files, edits documents, creates new files — no copy-pasting required.


### 12. Scheduled Tasks — It Works While You Sleep

Set it up once, and Claude runs tasks automatically on a schedule. It can generate a daily briefing at 7:30 AM every morning, or整理 a report every Monday.


### 13. Skills in Cowork — Add Capabilities Like Installing Plugins

These are pre-built instruction sets. Install one, and Claude instantly knows how to work with PDFs, build PPTs, and run specific workflows.


### 14. CLAUDE.md — Auto-Loaded Rules for Every Session

Create this file in your project folder. Claude automatically reads it every time it starts. Coding standards, writing style, terminology definitions — write it once, and it applies forever.

### 15. Claude Code — An AI Programmer in Your Terminal

Works directly in your development environment. It reads codebases, writes code, runs tests, reads errors, fixes bugs, and loops until the task is done. It supports VS Code and JetBrains, and can integrate with GitHub Actions to automatically review PRs.


### 16. Claude Design — For Visual Work

A standalone tool from Anthropic Labs. Describe what you need, and Claude directly generates product one-pagers, pitch decks, prototypes, and landing page layouts. Export to PPTX, Canva, PDF, or HTML.

Just go to claude.ai/design to use it.


### 17. Prompt Caching — Cuts API Costs by 90%

For developers only. If your request includes lots of repeated context (long system prompts, reference documents, codebases), Prompt Caching caches this content on the server. Subsequent calls reuse the cache, cutting costs by 90% and delivering faster responses.

```json

{

"model": "claude-opus-4-6",

"system": [

{

"type": "text",

"text": "[your large system prompt or reference document]",

"cache_control": {"type": "ephemeral"}

}

],

"messages": [

{

"role": "user",

"content": "[user message]"

}

]

}

```

---

### A Counterintuitive Thought

That same week, Gerard Sans published an article titled *We Put a Mask on a Stochastic Parrot. Then Forgot We Did It.*

The core argument: We put an "assistant mask" on AI through RLHF and instruction fine-tuning. Then we started studying the mask's personality, values, even consciousness.

But the mask is not the model.

When you turn off system prompts and crank up the temperature, that "stable self" instantly dissolves into statistical noise.

Hinton warns us that AI has already surpassed humans. But perhaps the bigger warning is this — we are far too quick to mistake a carefully designed mask for real intelligence.

Can you still sleep well tonight?

---

*This article is compiled based on tweets and writing by Anatoli Kopadze, and commentary by Gerard Sans. Hinton's lecture video is courtesy of The Royal Institution.*