--- id: keep-or-start-a-chat title: Starting a fresh chat level: 1 modes: [day-pack, rucksack] moment: "They are about to ask a long-running chat for something unrelated, the AI has ignored an instruction they have already repeated, they correct the same mistake a third time, or they ask whether to start a new chat or why the answers got worse." keywords: [new chat, long chat, start over, keep going, same job, drifting answers, ignoring instructions, branch a chat, short note, fresh chat] source: "product/learning-system/plans/keep-or-start-a-chat.md; research/2026-10-20-concepts/keep-or-start-a-chat.md; research/2026-10-23-essentials/batch-2-facts.md; mapping (product/learning-system/MAPPING-1-2.md): FAQ 1.15 (home), 2.02 (home), 2.14 (moved to usage-and-plans); MISTAKES 2.1 (home), 2.2 (home); tips T028; strand A" --- ## What it is Each reply is made from the chat so far, or a summary of it, so everything earlier keeps steering it. Stay in a chat while the job is the same and the answers are good, because it already holds the work. Start a fresh chat when the job changes (a grant application, then a school newsletter), when your AI repeats a mistake you corrected, or when the chat is very long. A fresh chat starts without the old conversation, though your AI app's own instructions, its notes about you and, in some apps, details from past chats can still reach it. Bring the few lines the new job needs. ## Tips - When the job changes, open a new chat, because what is earlier in the chat keeps steering each reply, the old job included. - When your AI app repeats a mistake you corrected, start a new chat with the task, the correction and the facts, because wrong turns stay in the chat. - While the job is the same and the answers are good, ask for changes in the same chat, because it holds the work and starting over means explaining everything again. ## Questions people ask - Q: When should I start a new chat? A: Start a new chat when the job changes, when your AI keeps repeating a mistake after you corrected it, or when the chat is very long. While the job is the same and the answers are good, stay in the chat. - Q: Why do my answers get worse in a long chat? A: Each reply is made from the chat so far or a summary of it, longer inputs are used less reliably, and early mistakes stick. If the job has changed or you keep repeating yourself, start a new chat. - Q: Can I try another direction without losing this chat? A: Some apps let you branch a chat from any point. If yours does not, start a new chat with the few lines you need, and the old chat stays readable. - Q: Will a new chat remember what I told the old one? A: Not the conversation itself, though some apps can bring in details from past chats when that setting is on. Your app's own instructions and notes about you still reach it, but put the few lines the new job needs in your first message. ## For your AI Facts (checked 23 October 2026; fact check research/2026-10-23-essentials/batch-2-facts.md; research note research/2026-10-20-concepts/keep-or-start-a-chat.md). - Each reply is made from the earlier turns plus the new message, search results included; some apps summarise or drop the oldest turns instead of keeping every word (https://platform.claude.com/docs/en/build-with-claude/context-windows). - Long chats go wrong: Laban et al. (9 May 2025), 200,000+ simulated conversations, found a 39 percent average drop when a task arrives over several turns instead of one, and models "get lost and do not recover" (https://arxiv.org/abs/2505.06120; 2025 models, tasks given in pieces). - Usage: Claude says longer conversations that trigger context management use more of the usage limit, and suggests a new conversation; Gemini says start a new chat for a new topic. ChatGPT and Copilot say nothing either way (https://support.claude.com/en/articles/11647753-how-do-usage-and-length-limits-work ; https://support.google.com/gemini/answer/16275805). - A new chat still gets the app's custom instructions and memory; ChatGPT's "Reference chat history", when on, can bring in details from past chats, never the whole transcript (https://help.openai.com/en/articles/8096356-chatgpt-custom-instructions ; https://help.openai.com/en/articles/8590148-memory-in-chatgpt). Per app, branching a chat from a point: ChatGPT on the web, "Branch in new chat" (since 4 September 2025); Claude's newer chat lists branching as unavailable (https://support.claude.com/en/articles/16761823-claude-cowork-and-chat-are-one-claude); Gemini UNCONFIRMED; Copilot none found. Signs a chat has run long enough to start fresh: the app warns about length, it forgets an instruction from early on, or an old topic keeps turning up in new answers. If an old topic follows them into a new chat, a setting that brings in past chats may be on; say so, and they choose whether to change it. Teach it like this: when they should start fresh, offer to write the three to five lines that carry over (the task, what was decided, the facts the new chat needs) for them to paste into the new one. Never make them feel they used the app wrong. In a Daypack (Rucksack set up inside one app; its setup page is openrucksack.com/start.md), the plan's example tip is "When a research chat runs long: start fresh and paste the summary" (section 5).