Make it yours: memory, Projects, repeatable workflows

Eight days in, you can brief, verify, and build. And yet every chat still begins with you introducing yourself — again. Today you stop re-introducing yourself to your own assistant: standing instructions it follows everywhere, a memory you can read and edit, and a workspace that already knows your files. This is the day the time savings stop being one-offs and start compounding.

1The silent time-leak: groundhog-day prompting

Think back over your last eight days of chats. The Briefing Formula email on Day 2. The Deep Research shortlist on Day 4. The week plan on Day 5, the deck outline on Day 6. Real wins, every one. Now notice what you did before each one: you introduced yourself. Who you are, what you do, what you're working on, how you like your answers. Again. And again. And again tomorrow.

Here's why. On Day 1 you learned that a chat remembers what was said within the thread — and that's exactly where it ends. A brand-new chat is a blank slate. Nothing carries over from yesterday's conversation unless the assistant has been given somewhere to keep it. That's partly by design: a fresh chat that knows nothing about you is also a chat that can't mix up your projects or leak one topic into another. But it means that, out of the box, every new conversation is a new intern's first morning on the job.

The cost is real and almost perfectly invisible. Call it groundhog-day prompting: re-typing your context into a fresh chat, forever. Each instance feels like "only" a minute or two — which is exactly why nobody notices it. But it repeats across every fresh chat, every day, week after week, and those small re-introductions quietly add up. The exact toll depends on how often you start from scratch, but for most people it's a meaningful slice of time spent telling the same tool the same facts about the same person — typed away a minute at a time, by someone who already learned how to brief on Day 2.

Day 1's Confident Intern frame had three consequences, and we've cashed in two: you brief an intern (Day 2), and you check an intern's work (Day 4). Today is the third: an intern gets better the longer they work with you — but only if they're allowed to keep notes, hold a standing brief, and have a desk where your files stay put. Today you hand your intern all three.

2The Persistence Ladder: three rungs of "it knows me"

Every major assistant — ChatGPT, Claude, Gemini — now offers the same three levels of persistence, under slightly different names. We'll call it the Persistence Ladder, and you'll climb all three rungs today. Each rung answers the same question — "what does the assistant already know when I start typing?" — with a bigger answer than the last.

Rung 1 — the thread. This is the cheapest persistence there is, and most people waste it. Because a conversation remembers itself, the move is to keep one ongoing thread per topic instead of scattering a topic across ten fresh chats. Renovating the kitchen? One thread, renamed "Kitchen renovation" (the Day 1 renaming habit), returned to every time. The context you built last Tuesday is still sitting there on Thursday. The limits are obvious: long threads get cluttered, and nothing in them helps a different chat. That's what the next rung is for.

Rung 2 — the account level. Two tools live here, and they're different in a useful way. Custom instructions are standing orders you write deliberately — a short note the assistant applies to every new chat: who you are, what you do, how you like your answers. Write them once, and you never again type "keep it short and skip the buzzwords." Memory is the same idea, but accumulated: as you chat, the assistant quietly retains facts about you — your job, your projects, your preferences — and uses them in later conversations. As of mid-2026, all three assistants include memory on free accounts — ChatGPT's free tier has a lighter version (a deeper memory upgrade is still rolling out as of June 2026), Claude added free memory in March 2026, and Gemini includes it. These plans shift often — if availability looks different on your screen, the workflow still works the same. The principle of rung 2 in five words: standing orders beat repeated orders.

Rung 3 — Projects and Gems. A Project (ChatGPT and Claude) or a Gem (Gemini) is a workspace: standing instructions plus your uploaded reference files, scoped to one area of your life. Every chat you start inside it begins already knowing the context — your CV, your client notes, your renovation quotes. Remember grounding from Day 5? When the assistant works from text you supplied, confident mistakes drop dramatically. A Project is grounding made permanent: instead of re-attaching the same files every time, they live in the workspace, and every chat draws on them automatically. As of mid-2026, Projects are available on the free plans of both ChatGPT and Claude (Claude's free plan caps how many you can have), and Gems are part of free Gemini. This rung is the single biggest quality-of-life upgrade in this course.

3Memory is yours: view it, edit it, delete it

A tool that remembers you across conversations deserves a clear-eyed minute, so let's give it one — calmly. The Privacy Line from Day 5 said: don't paste anything you wouldn't email to an outside contractor. That rule still stands, and it applies with full force to the files you upload into a Project. Today adds the second half of the rule, and it's good news: memory is not a black box. On all three assistants you can open the memory controls, read every fact that's been stored about you, edit what's wrong, and delete what you'd rather not have there — one fact at a time or all of it. The Privacy Line, revisited, isn't a warning. It's control.

There's a second, less obvious reason to look inside: stale facts. Memory accumulates whatever you talked about — including the job you left, the trip you already took, the diet you abandoned in February. An assistant working from outdated facts is like a well-meaning colleague who still thinks you're in your old role: not malicious, just confidently behind. So treat memory the way you'd treat that colleague — update it. A practical cadence: once a month or so, ask your assistant what it remembers about you (you'll do this in a minute), and spend ninety seconds pruning. Two questions per fact: Is it still true? and Do I want it in there? Delete accordingly, with zero guilt.

4From wins to workflows: your first playbooks

Persistence has one more layer, and it doesn't live in any setting — it lives in your notes. Since Day 5 you've been keeping a Prompt Notebook: every prompt that worked, saved verbatim. A recipe box, not a textbook. Today the notebook graduates. A prompt that worked once is a win; a prompt with a schedule and a quality bar is a workflow. We'll call that a playbook, and it has exactly three parts: trigger → prompt → done-check. The trigger says when you run it, the prompt is your saved winner with placeholders for the parts that change, and the done-check says how you know the result is good enough to use.

Here's the shape, using the most useful playbook in the course:

Three playbooks are worth assembling from the days you've already done — consider them the starter set:

• The Weekly Planning ritual — trigger: Monday morning. Prompt: your Day 5 Plan winner (above). Done-check: a realistic plan plus a "won't fit" list you've pushed back on at least once.
• The Inbox triage ritual — trigger: opening an inbox with more than five unanswered emails. Prompt: your Day 5 Draft winner ("Here's an email I received: […]. Draft a reply that […]"). Done-check: a reply you've read end-to-end and lightly edited before sending — the intern's draft never leaves your desk unread.
• The Learning ritual — trigger: a question that's come back three times this month. Prompt: the Day 4 research prompt for the big version, then Day 5 Distill on what comes back. Done-check: the Three-Click Check done, one summary saved where you'll find it.

One sentence on the horizon, then we move on: paid tiers of the big assistants now offer scheduled and automatic tasks that can run routines like these for you — genuinely nice later, genuinely not needed now. A playbook in your notebook plus a recurring slot in your calendar does the same job for $0.