For Jeremy Adas · April 2026
What it is. How every part of it works. How to use it well — written for a Chromebook, a curious mind, and someone tired of being scattered across four AI tabs.
You're already building things on Perplexity from your phone.
That's not "not technical." That's a builder who hasn't named himself one yet.
So this isn't a Claude tutorial — it's a map. The lay of the land, the surface area, and how to actually work with it. Read it cover to cover, then refer back whenever you hit a feature you don't recognize.
The map
Part 1 places Claude in your stack. Part 2 walks every feature, plain English. Part 3 is the craft — how to talk to it, the moves that matter, the traps to dodge.
What Claude is. Why people pick it. How it compares to Perplexity, GPT, and Copilot.
Jump in →Every Claude feature, in plain English — Projects, Artifacts, Files, Styles, Memory, Voice.
Jump in →Prompting craft, power moves, real workflows, common pitfalls, and a quick reference.
Jump in →Part 01 · The Lay of the Land
You don't have to pick Claude OR Perplexity. They do different jobs. Use both. Use Copilot at work, within Fidelity's policy. Stop scattering.
Primary · Personal
Thinking, writing, building. Web-based — runs perfectly on your Chromebook. Three model sizes (Opus / Sonnet / Haiku); Sonnet is the right default for almost everything.
Research · Personal
Keep using it. "What's the current state of X?" with citations. Computer-use agent for go-do-this-on-a-website tasks. Already strong for you.
Work-Only
Inside Fidelity policy. Don't mix work data into your personal Claude account. Different account, different rules — keep them separated.
Casual Fallback
Image generation is faster here. Light queries, fun stuff. Doesn't earn the primary slot — Claude wins on writing, reasoning, and building.
Claude vs ChatGPT
They look similar from the outside. The differences show up in use. Here's where each one actually wins — so you can stop wondering which to open.
Where Claude wins
Where ChatGPT wins
Part 02 · The Surface Area
Claude is mostly a website (claude.ai) plus a mobile app. Same conversations and projects sync across both. Sign in once, you're set. Here's everything you can do with it.
Feature · 01
The main website. Bookmark it. Where conversations, Projects, settings, and history live. Looks the same on a Chromebook as on a Mac.
Feature · 02
A single chat thread. Pick a model, type a message, get a response. Disposable — close the tab and (unless saved in a Project) it's gone.
Killer · 03
A persistent workspace. Name it, write a system prompt once ("you're my coding tutor"), upload reference files. Every new chat starts with that context.
Killer · 04
A side panel where Claude builds things you can preview live — HTML pages, mini-apps, code, formatted docs. Fastest way to build "fun apps" on a Chromebook.
Feature · 05
Drop in PDFs, images, screenshots, code, CSVs, Word docs. Claude reads them. Up to ~100 files in a Project. No conversion needed.
Feature · 06
Paste or upload an image and Claude can see it — diagrams, photos, screenshots, even handwritten notes. Useful for "what does this error mean."
Feature · 07
Set Claude's voice. Built-ins: Concise (short answers), Explanatory (teaches you), Formal. Or write your own.
Feature · 08
Claude remembers across chats — your preferences, recurring topics, things you've told it. Reviewable and editable in Settings. Off by default.
Feature · 09
Same Claude, on iOS or Android. Tap voice to speak instead of type — useful on a walk, in the car, when an idea hits. Transcripts save to the conversation.
Beta · 10
Claude can drive a browser/computer like Perplexity's agent — click, type, navigate. Powerful but rough. Don't start here; come back when you're comfortable with the basics.
Plans & pricing
Free is fine for trying it out. Pro is where you'll actually live. Max is for heavy users — probably overkill at the start.
Tier · 01
$0
Tier · 02
$20 / month
Tier · 03
$100–$200 / mo
Claude Code (CLI for developers) and the API have separate pricing — both out of scope for this guide. Start with Pro.
Part 03 · Working With It
The single biggest mindset shift: talk to Claude like a smart, eager intern — not a search engine. The intern needs context, examples, and feedback. Give it those and it's brilliant. Skip them and you'll be frustrated.
Set the scene before you ask the question. Who you are, what you're working on, what you're optimizing for. Two sentences of context beats a perfectly worded one-liner.
Bad: "Build me a budget app."
Good: "I'm building a personal budget tracker for myself.
I'm on a Chromebook. I want it to be a single HTML
file I can run in my browser. Build me v0."
"Make it casual" is vague. "Make it sound like this email I sent my brother" is specific. When you can, paste a sample of what good looks like — and a sample of what bad looks like.
Don't say "build the whole thing." Say "before you write any code, walk me through your plan." Read the plan. Push back. Then say "go." Catches misunderstandings before they cost you an hour.
When Claude misses, say what's wrong specifically inside the same chat. "That's too long, cut paragraph 2." "Tone is too formal, make it casual." Claude remembers everything in the conversation. Starting fresh loses that context.
A power move. "Here's what I'm trying to ask. How would you rewrite this prompt to get a better answer?" Claude is great at this. Use it whenever your output isn't quite landing.
"I'm not technical, explain in plain English" calibrates the response. "I'm a senior engineer, skip the basics" does the opposite. Claude can't read your mind — tell it where you're starting from.
Power moves
Once the rules feel natural, layer these on. They're the difference between using Claude and getting compounding value out of it.
Move · 01
Pick something recurring — a coding sandbox, a writing project, a research topic. Add a 3-sentence system prompt that says who Claude is and how to talk to you. Drop in a few reference files. Now every chat starts with full context.
Move · 02
Every time a prompt works really well, save it. Open a doc called "Claude Prompts." Paste in the prompt + a one-line note about when to use it. After a month: 20–30 reusable starts.
Move · 03
Perplexity researches with citations. Claude builds and writes. Workflow: ask Perplexity "what are the top 5 things known about X?" → paste into Claude → "use this to draft a brief / build a tool / explain to a 5-year-old."
Move · 04
You're already doing this on Perplexity. Try it in Claude. Ask for a single-HTML-file mini-app — a budget tracker, a habit log, a recipe organizer. Iterate in the Artifact panel. Save the HTML, run it locally.
Move · 05
An idea hits on a walk? Open the mobile app, tap voice, talk. Claude transcribes and responds. Capture before it evaporates. Refine on a real keyboard later.
Real workflows
Pick one. Run it tonight. The fastest way to learn Claude is to use it on real things you'd be doing anyway.
"I want to build a thing" → Artifact-driven build
New Project → describe it in 2-3 sentences (who it's for, what it does, what "done" looks like) → "Build me v0 in a single HTML file" → iterate in the Artifact panel until it works → save the file.
"I want to summarize a PDF" → Drop & ask
Drag the PDF into a new chat → "Summarize this in 5 bullets, then list the 3 most important quotes with page numbers." Claude reads it directly — no copy-paste, no chunking.
"I want to learn something" → Claude-as-tutor
Set Style to Explanatory → "Teach me [topic] like I'm intelligent but new to it. Ask me one question at a time and adjust depth based on my answers." A patient, infinitely-available tutor.
"I want to write something" → Co-writer mode
Tell Claude: who's it for, what's the goal, what tone, how long → ask for a draft → edit by saying what you'd change ("shorter intro," "less corporate," "add a story") — not by rewriting yourself.
"I want to plan a trip / project / event" → Constraint-driven planning
Drop the constraints (dates, budget, people, must-haves, can't-haves) → "Help me plan this. Ask me 5 questions before you start." The questions surface things you forgot.
"I want to send a hard email" → Diplomacy assistant
Paste the situation in plain language → "Help me write this email. Tone: direct but not aggressive. Audience: [who]. Goal: [what]. Show me 3 versions ranging from gentle to firm." Pick one. Edit. Send.
Common pitfalls
Everyone trips on these. Now you don't have to.
At a glance
Bookmark this page. Most "what tool do I open for this?" questions stop here.
| I want to… | Use… |
|---|---|
| Talk to Claude | claude.ai or the mobile app |
| Save context for next time | A Project |
| Build something visual / runnable | Artifacts |
| Read a PDF or doc | Drag & drop into chat |
| Show Claude an image or screenshot | Paste / upload it |
| Set Claude's tone | Styles (Concise / Explanatory / Formal) |
| Talk by voice | Mobile app · voice button |
| Find a fact with sources | Perplexity (not Claude) |
| Use AI inside Fidelity policy | Microsoft Copilot (not Claude) |
| Generate an image | ChatGPT (Claude can't yet) |
Things to leave at the door
These are real beliefs you walked in with. Each one is wrong in a way that's costing you time. Worth knowing before we start.
You're going to be good at this. The fact that you're already building "fun apps" on your phone tells me everything — you've got the instincts. Use Claude for a week. Tell me what you built.
Email Reu — let's go