I usually communicated, but suddenly when I sent an SMS, I got this text: I’m going to read the previous chat transcript to see what you were building and what (if anything) is still left to do, then I’ll either continue implementation or ask a focused question if the transcript ends with an explicit next step.
The transcript is large, so I’ll search within it for the last explicit request / “next step” and any open issues (e.g. “todo”, “bug”, “next”, “sound”, “mobile”, “deploy”).
Got it — the prior chat evolved from a Godot runner into an HTML canvas runner, and the todo list says coins/gaps/high score are already done. Next I’ll inspect your current workspace (игрыыы) to see what files exist and what state the game is in, then I can continue from there.
Briefly, it says that he read my 1st conversation and wants to continue, but how can I get my 1st conversation back? If there are no new chats and he just appeared in the specific chat where the conversation was active
@Danil_Simka thanks for raising this and for the clear context.
That message appears when a chat gets very long and the assistant runs a summarization pass over your existing conversation. It rereads the transcript and searches for the last clear request and any open “todo” items so it can continue the work in the same thread, but the wording is definitely confusing and we want to make this behavior feel less like a system log and more like a normal reply.
To make long‑running projects more reliable, it really helps to regularly ask the AI to write down the key findings, decisions, and important code snippets into separate .md files in your repo or workspace (for example: design-notes.md, api-contract.md, open-todos.md). That way, even if a chat eventually gets too long or you start a fresh chat, you still have a clean, versioned record that you can attach or reference again.
It also helps to keep each chat focused on a single task or feature and start new chats more frequently instead of putting everything into one huge conversation. That keeps token usage lower and makes it easier for the assistant to stay on track.
If you need to reuse work from an older chat, you can either:
Use the @Past Chats option (if it’s available in your build) to pull in a previous conversation, or
Open the old chat, export or copy the important parts (code, plans, notes), and paste or attach them into a fresh chat so the assistant has that context again.