You Do Not Need a Better AI. You Need Better Shortcuts.
You do not need a better AI. You need better shortcuts.
Look at your last week of chats and you will find the same sentences over and over. “Summarize this.” “Make this shorter.” “Find the holes in my argument.” Most people retype a handful of instructions dozens of times a week – slightly differently each time, so the answer comes back slightly differently each time too.
The fix is simple. Save each instruction once, give it a name, and call it with one word from then on. The reel behind this article collects ninety-nine of them across eight families. This article is about the part that matters more than the list: where a shortcut actually lives, how to write one that behaves the same every time, and how to choose the five worth saving first.
From retyped sentence to one-word command
Every shortcut follows the same path. The top row is the process; the bottom row is what it looks like for a real one.
| 1. Log → | 2. Count → | 3. Write once → | 4. Call → | 5. Review |
|---|---|---|---|---|
| One ordinary week Note every instruction you type |
Find the repeats Anything typed three times or more |
Name and define What it does, and what it must not |
One word Instead of a paragraph |
After two weeks Tighten, merge or delete |
| “Can you summarise this in a few bullet points?” | Typed 14 times, 14 different ways | Five bullets, most important first, no intro | /summarize | “Add one line: what to do next” |
The reel is honest about one thing that most lists leave out: these are not built-in Claude commands. They are shortcuts you set up yourself. That is not a limitation. It is exactly why they work – they describe your work, in your words, with your standards baked in.
Where a shortcut actually lives
There are two places to save one, and they behave differently. Pick based on where you do the work.
| Claude Code | Claude.ai | |
|---|---|---|
| Where it is saved | A Markdown file in .claude/commands/ inside the project, or in your personal ~/.claude/commands/ |
A short definition in the Project instructions |
| What it is | A real feature: the file name becomes the command | A convention the assistant follows because you defined it |
| How you call it | Type /summarize and pick it from the menu |
Type /summarize in any chat in that project |
| Passing extra detail | $ARGUMENTS in the file is replaced with what you type after the command |
Anything after the word is read as part of the request |
| Scope | That project, or everywhere for personal commands | Only chats inside that project |
| Sharing | Commit the folder and your team gets the same commands | Anyone added to the project sees the same instructions |
If you mostly work in the chat app, the project route is the one to use, and the ten-minute setup is in stop re-explaining yourself. If the words project, command and skill still blur together, twenty Claude terms, explained sorts them out.
What a command file looks like
In Claude Code a command is just a text file. This is the whole of .claude/commands/summarize.md:
Summarise the material below in exactly five bullet points.
Most important point first. One sentence per bullet.
No introduction, no closing remarks, no opinions of your own.
End with one line starting "Next step:".
Material: $ARGUMENTSNotice what is in it: a quantity (five), an order (most important first), exclusions (no introduction, no opinions) and a fixed ending. Those four details are what make the output identical on Monday and Friday. A command that only says “summarise this” is still a vague request – it just has a slash in front of it. We went through that difference line by line in slash commands are shortcuts you define, and it is the same constraints slot as in vague in, vague out.
The post this came from
Watch the reel
Ninety-nine shortcuts across eight families – swipe through and take the ones that match your work. Posted on the BANI Academy page.
Eight families, and one to start with in each
The reel groups its ninety-nine into eight families. You do not need one from each. But seeing them side by side makes it easier to spot which family your own repeats belong to.
| Family | A shortcut worth having | What it replaces | On a channel |
|---|---|---|---|
| Thinking | /holes | “Find the weak points in my argument” | Stress-testing a video’s main claim |
| Research | /sources | “Where does this come from, and how sure are you?” | Checking a fact before it goes on camera |
| Writing | /shorter | “Make this tighter without losing anything” | Cutting a script to length |
| Content | /hooks | “Give me opening lines for this” | The first ten seconds |
| Business | /reply | “Draft a polite answer to this email” | Sponsor and collaboration messages |
| Workflow | /checklist | “Turn this into steps I can follow” | Anything done three times by hand |
| Coding | /explain | “What does this code do, line by line?” | Small scripts and automations |
| Data | /trend | “What changed in these numbers, and why might it have?” | The monthly Studio export |
Two of these need a caution. /sources is only useful if you then open the sources; an assistant can describe a citation confidently and still be wrong about it. And /hooks gives you options, not a decision – the opening line still has to be one you would actually say, as we argued in draft five, record one.
Why you should not install all ninety-nine
A list of ninety-nine is a menu, not a setup. Save all of them and three things happen. You forget which exist, so you go back to typing. Several overlap, so you are never sure which to use. And a long block of definitions in project instructions competes for the assistant’s attention with the instructions that actually matter – your audience, your tone, your rules.
The better approach is the one in the diagram. Log a week, count the repeats, and save the top five. Those five will cover most of your day, because repeated instructions follow the usual pattern: a few account for most of the typing. After two weeks, look again. Keep what you used, merge what overlaps, delete what you never called.
This is the same discipline we apply to tool lists in we checked all eighty tools: a list is only useful once you have cut it down to what fits your actual work.
The quiet saving: fewer retries
The obvious benefit is typing less. The larger one is getting it right first time. When an instruction is written differently each time, the answer comes back differently each time, and a good share of chats become a loop of “no, shorter”, “no, without the intro”. Every one of those retries costs time, and on metered plans it costs credits.
A saved command removes most of that loop because the corrections are already inside it. That is the argument of you are paying for retries, applied at the smallest possible scale: one word, one reliable answer.
Frequently asked questions
Will /summarize work in a brand-new chat?
Only if it is defined somewhere that chat can see – a command file in Claude Code, or the instructions of the project the chat belongs to. Otherwise it is just a word.
Do these work in other assistants?
The idea does. Most capable assistants can follow a named instruction you define in their custom instructions. The file-based version is specific to tools that support it.
How many should I start with?
Five. Fewer than that and you will not build the habit; many more and you will not remember them.
What makes a definition reliable?
A quantity, an order, at least one thing it must not do, and a fixed ending. Four details, one or two sentences.
Save your first five this week
For the next five days, keep a note open and jot down every instruction you type into an assistant. On Friday, circle the five you typed most. Write each one properly – quantity, order, exclusion, ending – and save it where every chat can see it.
From then on, one word does the work of a paragraph. And the answer comes back the same way every time.
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 shortcut would you set up first? Tell us in the comments on the original reel, and we will show you how to write it.
For more on building channels for English-speaking audiences, that is what we work on at mmoyoutube.com.
One Comment