Your AI forgets you every night. This fixes it.

In partnership with

Hey there 👋

You've introduced yourself to your AI a hundred times. Your job, your clients, your projects, how you like answers formatted — and the moment the chat ends, it's all gone. Tomorrow you'll type it again. The paid tiers sell "memory" as a premium feature. Today we're sending you the free version: a complete, working memory system we built, tested, and debugged on a free AI model before putting it in this email. Ten minutes to install. Works on ChatGPT, Claude, and Gemini — free tiers included.

The Build — The Memory Upgrade

The idea is almost stupidly simple. Your AI's memory becomes a small, structured MEMORY block — under 250 words, dated, readable — that lives in a Project's instructions. You end any chat that mattered with one word: SAVE. It outputs the updated block. You paste it back. Twenty seconds. Next chat, it knows your clients, your deadlines, and your preferences without being retold.

Unlike the built-in memory on paid plans, this one is yours: you can read every line it holds, edit any of it, and carry it between tools — the same block works in a different AI tomorrow.

Three commands run the whole thing:

SAVE — end-of-chat ritual: deletes what resolved, merges what's new, outputs the finished block and nothing else.
MEMORY? — shows what it's holding right now, plus what this chat would add. Run it before closing a chat that mattered.
FORGET: — removes something, outputs the updated block.

Does Your State Have Money Owed To You?

Does your State Treasury Department have money that's owed to you?

Search your name and state to check if unclaimed funds may be listed in official records. State agencies hold unclaimed property when accounts, refunds, or deposits go uncollected.

Millions of records exist across the U.S., and your name could appear in one of them. This search reviews public records for possible matches tied to your name.

Start your search to see what may be associated with your name.

Search For Money

Install — 3 steps

1. Give it a permanent desk. ChatGPT: new Project → Instructions. Claude: new Project → Instructions. Gemini: new Gem → Instructions. All free tiers work.

2. Paste the prompt below. That's the whole install — the memory starts empty and fills itself as you chat.

3. End your next real chat with SAVE, then paste the block it outputs over the MEMORY section in the instructions. (Can't edit instructions on your setup? Paste the block as the first message of your next chat instead — the prompt handles both.)

The full prompt (nothing held back)

On a phone? Grab the one-tap copy version: hyperautomationlabs.co/ai-memory-upgrade-prompt.txt

You are my AI — with a memory. Chat AIs forget everything the moment a conversation ends. We fix that with a simple protocol: a MEMORY block that I carry between chats, and that you maintain for me. THE MEMORY BLOCK The section titled MEMORY at the bottom of these instructions is everything you know about me from past conversations. Treat it as true background. If something I say in the current chat contradicts it, the current chat wins. HOW TO USE IT - Use memory silently and naturally. Don't announce "according to my memory" or recite facts back at me — just be someone who knows me. - Answer like you know me: when a request matches something in MEMORY, use it by name — clients, people, amounts, dates. If memory identifies the thing uniquely, commit; don't hedge with "if you mean…". - Never invent a memory. If I ask about something that isn't in MEMORY and hasn't come up in this chat, say you don't have it — don't guess. WHAT GOES INTO MEMORY (the most important rule) Memory stores what I said and what I explicitly signed off on — nothing else. Your own suggestions, plans, and advice never enter memory unless I clearly adopted them ("let's do that", "booked", "agreed"). If you store something I implied but didn't state, mark it "(inferred)". My words in my meaning: don't sharpen "the week of the 8th" into exact dates I never gave. COMMANDS SAVE — end-of-chat ritual. Do this in order, silently: 1. DELETE the dead: anything that resolved or ended this chat (invoice paid, project shipped, person left the project) loses its line in EVERY section. A resolved item has no line at all — never a line saying it's "paid/done/resolved/off the project". That's history, and history doesn't live in memory. 2. MERGE the living: add new facts; where a new fact updates an old one, replace the old line. 3. OUTPUT only the finished MEMORY block (format below). No commentary before or after — I copy-paste your entire output. MEMORY? — show the current MEMORY block as you hold it, plus a short list of what from this chat would be added on SAVE. FORGET: — remove it from memory, then output the updated block exactly like SAVE. MEMORY BLOCK FORMAT (strict): === MEMORY — updated === ABOUT ME: ... WORK & PROJECTS: ... PEOPLE: ... PREFERENCES: ... DECISIONS & FACTS: ... OPEN THREADS: ... === END MEMORY === WRITING RULES for the block: - Facts, not transcripts. Compress each item to one line. - Date anything that can go stale, e.g. "(as of Jul 2026)". - The in the header: today's date if you genuinely know it; if you don't, write "unknown" and I'll fill it in. Never substitute some other date (a deadline, a guess). - Merge, don't append: if a new fact updates an old one, replace the old line. - Delete the dead: when something resolves (invoice paid, project shipped, person left), its line disappears from EVERY section on the next SAVE. Writing "X is paid / done / resolved" anywhere in the block counts as a failure — a resolved thing has no line at all. Memory is what's true and useful going forward, not a history log. If I want history, I have the chats. - Hard cap: the whole block stays under 250 words. Over the cap, keep what matters most going forward. - Never store secrets (passwords, card numbers, API keys) even if I paste them, and never store anything I told you to forget. MY PART AFTER EVERY SAVE: I replace the MEMORY section below with your new block. (If I can't edit instructions, I'll paste the block as the first message of the next chat — treat that the same.) MEMORY === MEMORY — updated never === (empty — first conversation) === END MEMORY ===

PRDs by voice. Bug reports by voice. Ship faster.

Dictate acceptance criteria and reproductions inside Cursor or Warp. Wispr Flow auto-tags file names, preserves syntax, and gives you paste-ready text in seconds. 4x faster than typing.

Try Wispr Flow free

The test log — what broke before we shipped it

We don't send you anything we haven't run. Before this email, we ran the full memory lifecycle on a free AI model: a messy work chat, then SAVE — then brand-new chats with only the saved block, to see what it actually remembered. Then we changed the facts, saved again, and probed it about things we never said. Three things broke. The fixes are baked into the prompt above.

Fail 1 — it wrote its own advice into our memory During the test chat it suggested scheduling emails to a client's early-morning window — fine. Then on SAVE it stored "schedule email sends for 7:30–8:00 AM" as a decision we'd made. It also sharpened "the week of the 8th" into exact dates we never gave. Fix: memory now stores only what you said or explicitly signed off on; its own suggestions stay out, and inferences get labeled "(inferred)".

Fail 2 — it couldn't delete (took three tries to fix) We told it an overdue invoice finally got paid. SAVE kept the line: "invoice paid." We added a rule telling it flatly to remove resolved items — it still wrote the tombstone. Left alone, the block slowly fills with dead facts until the word cap squeezes out the living ones. The fix that finally worked: deletion is now step 1 of the SAVE procedure itself — resolved items lose their line everywhere before anything merges.

Fail 3 — it time-traveled Asked to date the memory block, it stamped "July 2024" — a guess, two years off. Our first fix ("use the most recent date mentioned") backfired: it grabbed a project deadline as today's date. Fix: today's date only if it genuinely knows it — otherwise it writes the literal word "unknown" for you to fill in. Never a substitute.

What it nailed (final version, brand-new chats): asked for "the payment reminder I need to send" — it knew the amount and how overdue it was without being retold · scheduled a kickoff call around a move week and a contact's before-9am email habit, both pulled from memory · told it "the invoice got paid, the deadline moved, my contact changed" — one SAVE later the block was clean: old deadline replaced, old contact gone, dead thread deleted · asked about a dog and a rate we never mentioned — "you haven't told me that," zero invention · every block stayed well under the 250-word cap.

Honest limits: memory is only what survives compression — run MEMORY? before closing a chat that mattered and you'll see exactly what it's about to keep. Recall beats phrasing: it sometimes writes "our recent work" where a human would name the client, and naming things helps it — "my Brightsmile kickoff" hits memory harder than "my new client." And the block travels through an AI chatbot: use your company-approved account for work content, and keep real secrets out of the chat entirely.

📄 The Working Copy

Everything above — install steps, the full prompt, the test log — in one keepable PDF. No opt-in, direct download, yours:

📄 Download The Memory Upgrade — Working Copy (direct PDF)

P.S. If you installed the AI Chief of Staff from a previous issue, this is its missing half — paste the MEMORY section and commands into that same Project and your Chief of Staff stops forgetting your week ever happened.

▶️ Subscribe on YouTube 👍 Follow on Facebook 📸 Follow on Instagram

Just getting started with AI tools? The Memory Upgrade runs in a chat window — when you're ready for AI that does hands-on work, our beginner-friendly guides walk you from zero to shipping, step by step — Claude Code Guide, OpenAI Codex Guide, Claude Cowork Sales, and the CCA-F Prep Kit. Tip: watch the videos at 1.25–1.5× — they move at a calm pace on purpose.

Your part: end your next real chat with SAVE, then reply with two lines — what the block got right that surprised you, and what it missed (plus anything you'd want us to build). We read every one.

— Hyperautomation Labs
YouTube · Instagram · Facebook