Managing chats

What you'll walk away with: the cheapest tool on this site — a second window as a second opinion — and the handoff habit that keeps work alive across chats.

You need: both tabs, about 20 minutes, and a meeting agenda you need this week. The demo runs in your browser.

The chat list down the left of both tabs is history, not memory: each chat is a fresh context window with nothing of yours in it yet, and nothing from an old chat rides along unless something puts it there — you by pasting, or the app.

Step through it: every send re-packages the transcript — whole while it fits, trimmed or condensed by the app when it doesn't — and two weeks later there is nothing sitting there to come back to. Press SEND three times and read what each send carried: the preamble, every earlier message and reply labeled AGAIN, then the new one. Close the laptop — two weeks pass: the old tab is gone, and the window holds the preamble alone; no model was running. Paste the old script back in: now it "remembers", because you resent the script. The counts are illustrative — the toy's own face says SIMULATED — but the mechanism is the real thing.

The machine keeps no notes

One more consequence, and it is the one professionals miss: the machine keeps no notes. Whatever it weighed while writing an answer — the options you watched it hold and drop in the press-send x-ray, the paths not taken — is gone the instant the answer ends. Only what it actually wrote survives to ride back in on the next send. So when the reasoning matters, make it say the reasoning out loud. "Walk me through it before you answer" isn't politeness; it is how that reasoning becomes text in the window, where you and it can both use it next turn. One boundary keeps this discipline honest: what it writes down is a rendition of the work — more text from the same move, not a window into the machinery. So don't ask for a soliloquy; ask for the checkable parts: which sources it used, what it assumed, what it calculated. A written assumption can be challenged. A hidden one can't.

A fresh chat is a window with nothing of yours in it

A window that just built something is a poor judge of it. Its contents — your instructions, its own draft, every reason it already committed to — sit there tilting the odds toward yes, that's good. Ask it to find the flaws in its own work and you often get a polite note from a window with a stake in the answer — real flaws surface softer and rarer there. Nothing is being dishonest. You asked for the likely continuation of a window that already contains the thing being judged.

So don't. Open a new chat, paste in only the work — not the conversation that made it, not what you were hoping for — and ask what is wrong with it. A fresh window has no stake, and it drifts toward a different average because you loaded it differently. It is not a different mind, though — same model, same instincts, same blind spots. Treat what it raises as leads you resolve against the source, not as a second vote that settles the question. Build in one window, judge in another.

Naming, keeping, leaving

Name a chat for the task, not the date. Both apps let you rename a chat from the list (menus move; look for the nearest thing). Task names turn the list into a shelf — you will come looking; the model never will.

When a thread gets long, leave it. "It forgot what I told it" is almost never memory failing; it is the window ending — a new chat, or a thread long enough that the app quietly trimmed or condensed the middle to make room. The window is finite, and everything you add competes for it. Nothing broke. You started a new window and expected the old one to still be standing. So leave on purpose, carrying the work in three lines: what it is, where it stands, what was decided. Lesson 07 shows why a long thread goes dumb.

The chat is the script, not the result. Put the result somewhere it will live — a doc, a page, a file with a name: a source of truth a later window can read. The chat that produced it is a script of how you got there, not a place to keep it. Lesson 16 builds the home.

Do this in ChatGPT and Claude

Four chats, one tab first.

Chat A builds it. In a fresh chat, send:

Draft a five-bullet agenda for [a real meeting you have this week]. Ask me two questions first.

Answer its questions and take the agenda — or, if it drafted without asking, take the draft. Do not ask this chat to judge it.

Chat B is the second window. In a new chat, paste only the finished agenda — not the conversation that made it — with one line: What is wrong with this agenda? What would a careful chair challenge? Resolve each lead against the meeting you have; note what you decided.

Chat C is the lazy handoff. In a third fresh chat, send only Now add two more bullets. Predict first, then read: it has no idea what you mean — it asks what you are adding to, or invents something — because the window holds those five words, the preamble, and nothing of yours. (If it adds to your agenda, the app reached into your earlier chat — lesson 12 — and the model still read it off the window.)

Chat D is the handoff done. In a fourth fresh chat, paste three lines of your own — what the meeting is, the five bullets as they stand, what you decided in chat B — then the same five words. The new bullets should fit the meeting you described.

Compare C and D: the same machine, the same request; the difference is the window. Then the same four in the other tab, pasting the first tab's agenda into its chat B too — a second opinion from a different engine, which is why both tabs stay open.

---

Check your understanding: if no model is running between your messages, what is actually happening when a chat seems to remember last week? And why could chat C not know what you meant — or, if it seemed to, what did that?

Check yourself first — then open one good answer+−

No model runs between your messages, so a chat that "remembers last week" is the app loading the window well: the conversation rides in again on every send. Chat C was a fresh window with nothing of yours in it but five words — nothing to add bullets to; if it seemed to, the app had pasted for you. Chat D worked because the summary you pasted put the work back in the window; the handoff is the memory.

Your answer should have hit the window rebuilt each send, and the fresh chat as a window with nothing of yours in it until you paste.

Use this move whenever a thread gets long or a draft needs judging. Next: Today's date, the cutoff, and search.