Claude Projects: stop re-explaining yourself in every chat
Productivity

Claude Projects: stop re-explaining yourself in every chat

Claude Projects: stop re-explaining yourself in every chat

Imagine that every morning a brand-new colleague turns up at your desk. They are brilliant: they grasp anything the first time, they write fast, they never get tired and they never complain. But they have one serious problem, which is that by nine in the morning they have forgotten absolutely everything about the day before. They do not know who you are, or what you do for a living, or who your clients are, or why the Thursday newsletter is written in that particular tone and not another one. So you sit down with them, you go through it all again from the beginning, and by the time they are finally up to speed half an hour has disappeared.

That is exactly what you do every time you open a new conversation with Claude. You type three paragraphs setting out the context, you paste in the same document as always, you clarify that you write for people who are not technical, you remind it that you cannot stand grand, overblown adjectives. And the next day you do the whole thing again. If that sounds familiar, it is not that you are doing something wrong: it is that you are missing a piece. That piece is Claude Projects.

What a Project is, in plain terms

A Project is, above all, a work folder. But not just any folder: it is a folder with a note of instructions stapled to the outside, and a handful of reference documents kept inside. When you open a conversation inside that folder, Claude reads the note first and has the documents within reach. In other words, it starts talking to you already knowing who you are, how you like it to write, and what information you work with.

The difference between that and a loose chat is the same as the difference between calling a switchboard and calling the person who has handled your account for the past two years. To the switchboard you explain who you are every single time. To the person who knows you, you say "the usual, but for the new client" and you understand each other. A loose chat is the switchboard. A Project is the colleague who already knows you.

Inside a single Project you can open as many conversations as you like, and all of them start from the same context without you having to repeat it. One day you ask for a draft, another day for a review, another day you have a quick question. All three times, it knows what you are talking about. And the conversations stay grouped in one place, which helps when you want to go back to something you did three weeks ago and cannot remember where you left it.

Start with one specific thing that repeats

The most common mistake when people discover this feature is wanting to build a Project for everything. It does not work. A Project pays off when it covers a small, well-defined patch of ground that you keep coming back to.

Think of something in your own life that comes round every week or every month. It might be the newsletter you send to your subscribers, the lessons you prepare for one particular group, the household accounts you go through at the end of the month, the product descriptions for your shop, or the minutes of that club you are secretary of. Any of those makes a good first Project, because it has a stable context that you are currently writing out by hand over and over again.

If you are torn between several, pick the one where you have most often heard yourself saying "hang on, let me explain". That is where you will feel the difference.

The instructions: write them as you would for someone new

This is the part most people skip, and it is the part that makes the biggest difference. The Project instructions are that note stapled to the folder, and the best way to write them is to imagine that tomorrow somebody is coming in to cover your job and you have a single sheet of paper to tell them how things are done around here.

It does not have to look polished or sound professional. It has to be useful and specific. Write it in your own words, in ordinary sentences, as if you were telling them out loud over a coffee. Four things are worth including:

  • Who you are and who you are addressing, described precisely enough that you can picture the person who reads or buys from you.
  • What the typical job inside that folder is, meaning what you will be asking for most of the time.
  • How you want it to sound: warm or formal, long or short, with examples or straight to the point, which words you always use and which ones you cannot bear.
  • What it must never do, which is sometimes worth more than everything above put together.
That last point is pure gold. Lines like "do not use exclamation marks", "do not promise results I cannot guarantee" or "do not give me the answer as a list unless I ask for one" fix in a single stroke the habits that wear you down most. And because they live on the note attached to the folder, they apply to every conversation without you repeating them.

One practical tip for getting them right: write a quick first version, work with it for a few days, and keep track of what you have to correct by hand each time. Whatever you keep correcting is precisely what your instructions are missing. You add it, you try again, and after two or three rounds you have a note that genuinely works.

The documents: two or three that really matter

As well as the note, the folder holds reference documents. The temptation here is to empty your entire hard drive into it, and that is exactly what you should not do.

Upload what a new colleague would need on the desk in front of them on their first day. For a newsletter, that might be two past issues that represent your style well and a list of the topics you have already covered. For lessons, the syllabus and one example of a well-built exercise. For the household accounts, last year's summary and the sheet where you note down the categories you use. Documents that explain how things work in your particular case, not everything that exists on the subject.

The more irrelevant material you put in, the worse the answers get. And it makes sense: if you hand someone three key documents, they read them and work from them; if you dump forty folders on them, they get lost and end up going with the first thing they find, which is often not the most important. Something similar happens with Claude. Less material, better chosen, almost always wins.

And from there, you simply get on with it

Once the note is written and the documents are in place, day-to-day use becomes surprisingly boring, in the best sense of the word. You go into the folder, open a conversation and type "draft this week's issue, the topic is this". No introductions, nothing pasted in, no reminder about tone or audience. It already knows.

The first time you notice it, you feel a small sense of relief. By the fifth time, you realise how much time you had been giving away just explaining yourself.

What a Project does not do

It is worth being honest, because this is not magic. A Project does not give Claude endless memory of everything the two of you have done together: it gives it a starting context, which is a different thing. Nor does it learn on its own from your corrections. If you want something to change permanently, you have to put it in the instructions; fixing it inside one chat works for that chat and not much further.

It also makes no sense for one-off tasks. If you need to translate a stray paragraph or summarise an article that happened to land in your inbox, open a normal conversation and get on with your day. Building a folder with instructions and documents for something you will do once is wasted effort.

And there is one last thing almost nobody takes into account: instructions age. Your audience changes, you adjust your tone, you stop offering that one service, you move into a new phase. If you never revisit them, one day you find yourself arguing with the tool because it keeps insisting on something you told it months ago that is no longer true. Open them every few weeks and give them five minutes. It is cheap maintenance and it saves you a fair amount of frustration.

The real leap is not writing better, it is writing less

There is a widespread idea that making progress with artificial intelligence means learning to write longer and more elaborate instructions. It works for a while, and then it turns into a burden: you end up maintaining a file full of blocks of text you copy and paste, like someone dragging a suitcase everywhere just in case.

The real leap from beginner to advanced user goes in exactly the opposite direction. It consists of taking the context out of your fingers and putting it into the tool, so that you stop writing it. When your Project is set up well, your messages become short, almost telegraphic, because everything that matters is already stored in there. You write less and you get more, which is the most reliable sign that you are doing it right.

If you would like to build yours calmly and see real examples step by step, the Claude Projects course walks you through it from scratch. And if you would rather start on your own right now, pick that one task you have been explaining over and over for months, and finally give it a folder of its own.

Ready to learn AI?

Sign up free and access 2 Fundamentals courses. No credit card required.

Create Free Account
Share this article