Which tool to choose
To not lose the thread after a block, the method counts more than the tool, but some help more.
- If the AI lets you go back to the conversation from the history, after the reset you pick up exactly where you were: the thread is already there, you just have to write again.
- If you have to change AI because the first is blocked, the thread doesn't transfer by itself: you carry it, with a summary. Here the quality of the summary makes all the difference.
The move that works everywhere, and that's worth making a habit, is having the summary given to you before the block, not after. When you're already blocked, you often can't ask that conversation anything anymore.
How to do it
From a computer or a phone the principle is identical: capture the state of the work while you can still write.
- When you feel the conversation is long and the limit is approaching, stop for a moment and ask for a status update. The operating syntax:
We're about to be cut off. Give me an operational summary to pick up
afterward: what we're doing, the decisions we've made, the style
or rules we have to respect, and what the next step is.
Write it so that another AI can start from scratch by reading it.
- Copy that summary and save it outside the chat: in a note, in a document, wherever you like. It's your lifeline.
- When you resume on the same AI after the reset, if you find the conversation again in the history, reopen it and continue: the context is already there. If you can't find it, paste the summary.
- When you resume on another AI, paste the summary as the first message and add "let's pick up from here, continue the work".
- Check: ask a control question about the previous work ("recap for me in one line what we're doing and what the next thing to do is"). If it answers coherently, the thread is picked up; if it's confused, enrich the summary with the details that are missing.
A concrete example
Lucia is building with the AI the itinerary of an eight-day trip: she's already defined stops, budget and pace, and we're on the sixth day. She knows the conversation is very long and fears the block. Before continuing she asks for the operational summary.
The AI writes her a status update: chosen destinations, daily budget, preference for public transport, days still to plan. Lucia saves it in a note. When the limit kicks in shortly after, she opens another AI, pastes the summary and continues from the seventh day without repeating anything that was decided. Eight days of planning aren't lost to a limit: they all fit in ten lines of summary.
When it does NOT work (and how to fix it)
If you got blocked without having asked for the summary first
It's the awkward case. If the conversation is still visible (read-only), scroll through it and reconstruct the state by hand, copying the key decisions. If it's truly inaccessible, reconstruct the fixed points from memory and give them to the new AI. Lesson for next time: the summary is asked for before, not after.
If the summary is too vague and the new AI starts off badly
A generic summary makes for a bad restart. Enrich it with concrete details: not "we're writing a text" but "we're writing a complaint letter to a phone company, firm but polite tone, we've already cited the contract and the date". The more specific the summary, the more precise the resumption.
If you also have to transfer a file, not just the discussion
The summary carries the decisions, not the files. If the work was based on an uploaded document, re-upload it in the new conversation along with the summary. The summary says "what we're doing with this file", the re-uploaded file gives back the content to work on.
If the new AI changes style or tone from before
Every AI has its own voice. If you need the tone to be consistent, put it in the summary as an explicit instruction: "keep a formal and concise tone, short sentences, no exclamations". The style doesn't transfer by itself between tools: it has to be declared, and once declared the AI follows it.
A tip from someone who actually uses it
For long jobs, don't wait until you're near the limit to ask for the summary: have it given to you in stages, every time you close an important piece. That updated summary isn't only useful against blocks: it's also your working document, the record of where you are. When the AI is cut off, you already have in hand everything you need to start again elsewhere, and the block becomes a thirty-second pause instead of work to redo.
Frequently asked questions
Does the conversation get deleted when I reach the limit?
No. The limit blocks the sending of new messages, it doesn't delete what's there. The conversation stays in the history: after the reset you reopen it and continue from where you were. The limit is a pause, not a destruction of what's been done.
Can I have the new AI read the whole old conversation?
In theory yes, by copying everything, but it's awkward and often too long to paste. A good operational summary is more effective than the whole conversation: it contains the decisions and skips the dead ends. Paste the full dialogue only if it's short; for long conversations, the summary wins.
How often is it worth saving a summary?
For a job that lasts, every time you finish a phase or make an important decision. It isn't needed at every message: it's needed at the points where, if the conversation disappeared now, you'd want not to have lost that progress. Those moments are your natural save points.
Isn't it the sign that the AI doesn't really have a memory?
It's true, and it's worth knowing instead of discovering it at the wrong moment. Most AIs don't really carry the memory from one conversation to another: every new chat is born with a blank slate. That's why the summary you carry with you isn't a fallback for moments of block: it's the way the continuity of your work stops depending on the AI's memory and starts depending on you. The reliable memory, from one session to the next, is the one you keep yourself.