A free Claude account, a job description, two context documents, and a two-minute ritual. That's the whole afternoon. By the end of it, you'll open a conversation and the assistant will already know who you are, what you're working on, and how you like to work — instead of starting from nothing, every single time.
Right now, every conversation with an AI chatbot starts from zero. It doesn't know your situation, your projects, or how you like things done — you re-explain, every time, and it forgets the moment you close the tab.
Level 0 fixes that with nothing but the free account you can create in the next five minutes: a dedicated workspace, a written job description, two short documents about you and your work, and a small ritual that carries context from one session to the next. No database, no code, no monthly bill.
The result isn't a smarter AI — it's the same AI, given somewhere to keep what it learns about you. That's the entire trick, and it's the foundation everything else in this Guide builds on.
Go to claude.ai and sign up. All it asks for is an email address or a Google account — no card, no trial that quietly turns into a subscription. You'll be talking to it inside a couple of minutes.
Worth knowing before you start, because it shapes what you can build at this level: the free tier already includes everything Level 0 needs. You get Projects (a workspace that holds standing instructions and files — the whole point of this page), file uploads, Artifacts (documents and other content the assistant can produce and revise alongside the conversation), web search, a persistent Memory feature, and your full conversation history. What the free tier doesn't include — a coding-focused tool, the very top-tier models, and a deep-research feature — none of it matters here. Level 0 runs entirely on the ordinary chat experience.
One practical note: free usage works on a rolling window — a stretch of hours during which you can send a certain number of messages before you need to wait for it to reset. The app tells you when you're getting close and tells you when the window opens back up. You don't need to track this yourself; Step 7 below has a way to work comfortably inside it.
Look for the option to create a Project — the app may label the button slightly differently depending on when you're reading this, but the concept is the same everywhere: a dedicated folder for one area of your life, distinct from the general chat window.
Here's what makes a Project different from just chatting: anything you attach to it — instructions, documents — is automatically present in every conversation you have inside that folder, without you pasting it in each time. Think of it as the difference between hiring someone fresh off the street for every task versus having one assistant who already has a filing cabinet with your name on it.
Create one Project for this purpose. Name it something you'll recognize immediately — "Chief of Staff," "My Assistant," whatever feels natural. Every conversation from here on happens inside it, not in the general chat window.
Every Project has a place for custom instructions — a standing brief the assistant reads before every conversation in that Project. This is the single highest-leverage thing you'll write today: it's the difference between an assistant that improvises a new approach every time and one that behaves the same reliable way every session.
Find the Project's settings and look for custom instructions, then paste in something like this and adjust the details to sound like you:
Nothing here is technical — it's just written clearly enough that the assistant can't misread it. Tone, standing behaviors, and boundaries, all in one place, read before every conversation you have in this Project.
The job description tells the assistant how to behave. These two documents tell it what it's working with — the standing facts it would otherwise have to ask you to repeat every time. Write them as plain text or in whatever document editor you like, then upload both as files attached to the Project (look for an option to add files or knowledge to the Project — same idea as the custom instructions, just as attachments instead of a settings field).
Who you are and what your life currently looks like — the background that makes your other requests make sense.
What's actually in motion — the living inventory of projects, goals, and open questions. This one goes stale fastest, and that's expected; you'll update it as part of the ritual in the next step.
Don't over-engineer either document on the first pass. A few honest lines beats a polished paragraph you'll be too precious to edit later. You'll revise both constantly — that's the point.
This is the piece that actually delivers on the promise. Documents alone go stale; a ritual keeps them current.
Opening a session: start with something like "Read my context docs, then ask me what's changed since the last note." The assistant reads both documents plus whatever handoff note is sitting at the bottom of Current Work, and instead of you re-explaining where things stand, it asks you the two or three things it actually needs to know.
Closing a session: before you leave, ask it to "write a short handoff note — what we covered, what's decided, what's still open." Copy that note and paste it into the Latest Handoff section of your Current Work document, replacing whatever was there before. Re-upload the file (or edit it in place, if your document tool is attached live) so the next session starts from it.
Be honest with yourself about what this is: a small piece of manual ferrying, copying a note from a chat window into a document and re-uploading it. That's Level 0's one real friction, and it's deliberate — it costs you nothing to set up today, and it's precisely the itch that Level 1 exists to scratch, once you're ready to let the assistant read and write that memory directly. Don't solve it early. Feel it for a week or two first.
Setup takes an afternoon. Here's how to spend the week that follows so the habit actually sticks.
On the usage window from Step 1: the easiest way to live inside it comfortably is to batch your real work into one or two sittings a day rather than trickling messages across the whole day, and to notice roughly when your window resets so you can plan a long task for right after it, when you've got the most room. If you do bump into the limit mid-task, that's what the handoff note is for — end the session properly, and you lose nothing by waiting it out.
Level 0 isn't built for any one kind of work — it's a general-purpose setup, and it earns its keep on exactly the situations where re-explaining your situation every time would otherwise eat the whole benefit:
Current Work tracks the timeline, the moving-day logistics, and which boxes of decisions (new place, movers, address changes) are still open — so every conversation about it picks up mid-thread instead of starting with "so, we're moving, and here's why."
Personal Context holds how you like invoices phrased and which clients need extra care; Current Work holds what's outstanding. Ask for a status email and it already knows the account, the tone, and the history.
Personal Context carries your background and what you're looking for; Current Work tracks applications in flight and where each one stands. Draft a cover letter and it's already calibrated to you, not generic.
Current Work holds which rooms, which contractors, which decisions are pending (tile? paint color?). Ask "what's still undecided" and get a real answer instead of a blank stare.
Nothing in that exchange required a database, a plugin, or any technical setup — just the documents actually being read and actually being updated.
You'll hit the edges of Level 0 honestly and probably soon: the handoff-note copying starts to feel like a chore, or you want the assistant to update something the moment you mention it rather than waiting for you to paste a note in later, or you want it working from somewhere other than the one device you set this up on.
That itch is the whole point — it's what Level 1 exists to scratch: the same setup, plus a small free filing cabinet the assistant can read from and write to directly, so the ferrying stops and the checking begins. There's no rush to get there. Level 0, used properly, is a complete and genuinely useful system on its own — plenty of people are well served staying right here.