Intermediate
6 min read
Updated Aug 25, 2026
The Six Decisions That Cover (Almost) Everything In Claude
Most people use Claude the same way they used the first search engine they ever tried: one box, one question, and whatever happens to load. That works fine for a while. But Claude is really a small stack of separate decisions - where you open it, which model answers you, how hard that model thinks before it replies, how you structure the work, what you get handed back, and what it's connected to. Once you can see those six decisions on their own, using Claude well stops being a matter of luck.
This guide walks through all six in plain terms. Skim it once top to bottom, then keep it around as a reference for whenever you're not sure which lever to pull.
The six-level framework
01 Where you open it
04 How you work
02 Which brain you pick
05 What you get back
03 How hard it thinks
06 What it's plugged into
It's the same Claude underneath, but there are four different doors in, and each one is built for a different kind of moment.
Claude in the browser
claude.ai. Nothing to install, which makes it the one that works even on a locked-down work laptop.
Claude in Chrome
Reads whatever tab you're on, clicks around, and fills in forms on your behalf.
Claude on mobile
iOS and Android, with the same account and the same chat history, wherever you happen to be.
Claude on desktop
Works alongside your real files, folders, and apps. The only version built to reach directly into your machine.
Quick pick: research on the go, use the browser or mobile app. The same web form over and over, use Chrome. Working directly with local files, use desktop.
Five models, running from fast and light up to genuinely hard thinking. A bigger model isn't automatically the right call - it's usually slower, and it's overkill for anything simple. The goal is matching, not maxing out.
Claude Haiku
Fast and light. Built for high volume, simple jobs where speed matters more than depth.
Claude Sonnet
The daily driver. A solid balance of speed and intelligence for most everyday work.
Claude Opus 5
Built for hard thinking: judgment calls, deep analysis, and genuinely messy problems.
Claude Fable 5
The frontier tier. Reach for it when a task is genuinely difficult and it's worth the extra time.
Claude Mythos
Currently restricted to select organizations, not public yet.
Quick pick: quick answers or simple formatting, use Haiku. Everyday work, use Sonnet. A high-stakes decision or a hard problem, use Opus 5 or Fable 5.
Separate from which model you pick, there's a second dial: how much effort it spends reasoning before it gives you an answer. Four settings, and none of them is automatically correct.
Low effort
Formatting, extraction, cleanup. Same answer, delivered faster.
Medium effort
Everyday work: drafting, summarizing, editing.
High effort
The usual default. Good for analysis and decisions that actually matter.
Max effort
Save it for when a task genuinely deserves the extra wait.
The trap to know about: more thinking doesn't mean more correct. Long reasoning just gives a model more room to be confidently wrong. Match the effort to the task instead of defaulting to max every time.
The same conversation window can be used five different ways, and most people only ever discover the first one.
Chat
One question, one answer. Where most people stay, and there's nothing wrong with that.
Projects
Load your files and instructions once, and every new chat inside it starts already up to speed.
Cowork
Hand over the whole task and step away. Built for multi-step work that runs on its own.
Claude Code
Built for developers, and also the fastest way to get Claude to actually run something rather than just describe it.
Claude Design
A canvas for slides, prototypes, and visuals that you shape just by talking.
It also works inside tools you already use: Chrome, Excel, PowerPoint, and Slack.
The output doesn't have to be plain text sitting in a chat bubble. Depending on what you ask for, Claude can hand you something you can act on directly.
Artifacts
Not a wall of text. A document, dashboard, or deck that you can open and edit directly.
Real files
Word, Excel, PowerPoint, or PDF, ready to download and send exactly as-is.
Live artifacts
Dashboards that refresh with your real data every time you open them again.
6. What it's plugged into
The last piece is context: three ways Claude keeps up with your world instead of starting over from zero in every single chat.
Connectors
Gmail, Drive, Calendar, Notion, Slack, HubSpot, and dozens more. Claude reads and acts on your real accounts.
Memory
Carries what matters between chats, so you stop re-explaining yourself every single time.
Skills
Instructions you write once, and Claude follows them forever, in every chat after that.
Try it instead of reading it: if you've got a specific task in mind, our free
Which Claude Setup Fits My Task? tool turns the first and fourth decisions above into a 30-second quiz.
That's the whole map
Where you open it, which brain answers you, how hard it thinks, how you work, what you get back, and what it's plugged into. Six questions, and once you can answer all six without thinking too hard about it, you're not just using Claude - you're using it the way it's actually meant to be used.