Ten Claude Features, in the Order That Actually Pays Off
Most people use a fraction of what an AI assistant can do. If you are only asking it to write an email, that is the whole of it – and a list of ten features is the usual response.
A list is not the fix, though. Ten names with no order produces the same feeling as an app store: everything looks useful, nothing gets adopted. What helps is knowing which feature solves which complaint, and in what order they pay off.
The order that pays off
Four stages. Each one removes a specific annoyance, and each is worth doing before the next.
| Stage 1 · Context → | Stage 2 · Instructions → | Stage 3 · Build → | Stage 4 · Access |
|---|---|---|---|
| It knows your work A project holding your channel, your rules, your past scripts |
It knows your standards Saved instructions that survive every new chat |
It makes things, not paragraphs A page, a tracker, a calculator you can open |
It reaches your material Connectors, a browser, a desktop worker – read-only first |
| Feature: Projects |
Feature: Skills |
Feature: Artifacts |
Feature: Connectors · Chrome · Cowork |
| Ten minutes, pays back the same week | Half an hour per skill | Minutes, no setup | A permission decision, not a feature |
Stage one first, always. Everything else works better once the assistant already knows who you are writing for and what you refuse to do – the ten-minute setup in stop re-explaining yourself. Without it you are re-briefing a new colleague every morning, and no other feature on the list compensates for that.
All ten features, with the complaint each one fixes
Here is the full roadmap, reordered by usefulness rather than by name.
| Feature | What it is | The complaint it fixes | Setup cost | When you need it |
|---|---|---|---|---|
| Chat | The ordinary conversation | Everything that starts as a question | None | Day one |
| Projects | A workspace holding your context and files | Re-explaining your channel every session | Ten minutes | Week one |
| Skills | Saved instructions for a job you repeat | Answers that drift between attempts | Half an hour each | Week two |
| Artifacts | It builds a page, tool or document you can open | Wanting a thing rather than a description of a thing | None | When you need one |
| Connectors | It reads your files or accounts directly | Pasting the same data in every week | A permission decision | After stages 1–2 |
| Extended Thinking | More reasoning before it answers | Hard problems where the first answer is shallow | None | When the question is genuinely hard |
| Claude in Chrome | It works in your browser alongside you | Repetitive browser chores | Permissions, narrow scope | Later |
| Cowork | Desktop automation, no code needed | Files, PDFs, repeated busywork | Early beta – test on copies | Later |
| Claude Code | An agent that builds and ships in a codebase | Small tools, scripts, sites | Only if you code | If you build |
| Claude API | Programmatic access for developers | Putting it inside your own product | Developer work | Rarely, for creators |
Read the last column and the pressure comes off. Four of these ten are for people writing software, running automations or building products. If that is not you, skipping them is not a gap in your knowledge – it is a correct decision, the same one we made about enterprise rows in the wall is the signal, not the chart.
The post this came from
Watch the reel
The full roadmap of features on one card, with what each is for. Posted on the BANI Academy page.
Start from the symptom, not the feature
The fastest way through a roadmap like this is to ignore it until something annoys you, then look up the annoyance.
| What you catch yourself saying | The feature | First move |
|---|---|---|
| “I explain my channel again every morning” | Projects | Put the audience, the rules and three past scripts in one place |
| “The same request gives me a different shape every time” | Skills | Write the instruction once, with a quantity and an exclusion |
| “I want a tracker, not a description of a tracker” | Artifacts | Ask for the thing itself and iterate on it |
| “I paste the same numbers in every month” | Connectors | Connect one sheet or folder, read-only |
| “The first answer is always too shallow” | Extended Thinking | Turn it on for the decision, not for the email |
| “I do the same twenty clicks every week” | Chrome or Cowork | Automate one chore, watch the first three runs |
| “I need a small script or page built” | Claude Code | Only if you can read what it wrote |
Rows two and three are the two most creators feel every week. Drifting answers are almost always a drifting instruction rather than a weak model, which is the whole argument of thirty-one retypes; and once a saved instruction is used weekly to a fixed standard, promoting it properly is the step in from shortcut to skill.
If the vocabulary is the part slowing you down – project, skill, artifact, connector – the plain-English version of all of it is in twenty Claude terms, explained.
Why ten per cent is usually the right amount
“Most people use ten per cent of it” is meant to sound like a warning. Read another way, it is what healthy tool use looks like. Nobody uses ten per cent of a word processor and feels guilty about it; they use the part that writes documents and ignore mail merge until the day they need mail merge.
The difference with AI tools is that the unused ninety per cent is marketed at you constantly, so ignoring it feels like falling behind rather than like focus. It is not. The creator who has set up one project and two saved instructions, and uses them every single day, is getting more out of the tool than someone who has tried all ten features once.
The honest test is not how many features you have touched. It is whether the work left your desk faster this month than last month. If the answer is yes, the other features can wait until a specific annoyance sends you looking for them – which is exactly what a roadmap is for.
The three features that need care, not enthusiasm
Six of the ten are ordinary software you switch on. Three – connectors, the browser and the desktop worker – are different, because they act on real accounts and real files.
The sensible defaults are the same for all three: grant the narrowest access that does the job, start read-only, treat anything read as data rather than instructions, and keep anything involving money entirely out of scope. Those rules, and why they matter more as more of your week flows through one connection, are in stop pasting and same brain, three doors.
The fourth to treat carefully is not on the risk list but on the honesty list: a feature that produces a polished artifact makes weak thinking look finished. Keep the upstream questions where they belong, as in most people use AI to draft.
Frequently asked questions
Do I need a paid plan for all of this?
Availability differs by plan and changes over time. Check the current pricing page rather than a chart, including ours.
Is extended thinking always better?
No. It helps on genuinely hard problems and wastes time on simple ones. Use it for decisions, not for rewrites.
Which feature should a creator learn first?
Projects, then one skill. Together they remove the two most common daily annoyances and take under an hour.
What about the ones that act on my behalf?
Treat them as permissions, not features: narrow scope, read-only first, nothing near money, and watch the first runs.
Set up stage one today
Open a project. Put in three things: who your videos are for, the rules you never break, and two scripts you were happy with. That is stage one, and it takes ten minutes.
Next week, write one skill for the job you do most. Then stop, and only move to stage three when something in your week actually asks for it. A roadmap is not a checklist – you are not behind for skipping the features you do not need.
Learning a lot does not make anyone good. Doing a lot does.
If you want the structure that turns that into a habit, that is what our free challenge is built for: no barrier to entry, a genuine commitment to action, and a refundable commitment fee you get back when you finish. You are not paying for the knowledge. You are betting on yourself finishing, and we hold the stake.
Which feature have you never opened? Tell us in the comments on the original reel.
For more on building channels for English-speaking audiences, that is what we work on at mmoyoutube.com.
