The problem is, ChatGPT wasn't designed to be a proper writing or word processor tool. It treats messages as disposable, not like saved documents you can return to, structure, or modify easily.
Our notes of that time are amazing and flawless and better than the teaching material itself.
Teachers didn't like it as it benefited those that didn't listen at all, but for those of us who cared this was a major productivity boost.
Imagining an AI as additional user is crazy.
I used codex a few weeks ago to do some agentic coding. That actually is a model that could work well for working on groups of documents as well. Treating documents as code here and working to create coherent change sets against those would be a much better way to work with LLMs on writing. I haven't had a chance to try this yet. But using codex to edit a wiki might be the power move here.
It syncs your entire ChatGPT history to local markdown files so you can organize, rename, or integrate them with other work however you like. I use it to keep my conversations alongside related notes in Obsidian, but since it's just markdown, you can use any system at all.
I also built it because ChatGPT is a cloud-based service which, as we all know, can vanish, change its UI, or alter its terms at any time, and I wanted control over my own data.
Info, screenshots, & download all at https://martiansoftware.com/chatkeeper for the curious.
(full disclosure: I sell it, but it's cheap, not subscription-based, and has a free trial.)
(nerd disclosure: Reverse-engineering the ChatGPT export format is an ongoing game of whack-a-mole as they continually add new features. The software is currently a CLI but a GUI version is in the works.)
I will seriously consider that - thanks for the suggestion!
IMHO this is something to lean into. Students should not save chats like references. They should document their findings from the conversations and make sure they can ask again, should the need arrise.
The prompt they use to export the ChatGPT messages is:
---
Turn the following text into JSON format. The JSON should have three fields: "title", "prompt", and "data". The "title" should be self-generated, based on the main idea or theme of the input text. The "prompt" should be the exact prompt used to generate the "data" section. The "data" should be in markdown format, and it must include the original prompt as a section inside the markdown, followed by the input text. Keep any formatting like headings, bold, italics, and lists.
---
I asked them to include the prompt to preserve context so others can understand their thinking process and the goal behind the message.
Under that view, it is reasonable to save links to applications.
The core issue is that ChatGPT treats each conversation like a throwaway. Even if the content is valuable, it’s not saved like a document — there’s no real sense of “this is my project on climate change” or “this is the story I’m working on.” It’s just one long chat thread, and everything lives in a running timeline. You can rename a chat, sure, but that doesn’t make it feel like a proper file. There’s no version history, no folders, no tags, and no way to mark drafts or track progress. So students end up scrolling endlessly, trying to remember where they talked about that one idea or how they phrased something two weeks ago.-
Some of them are getting creative with workarounds. They paste stuff into apps like Notion or Docs, or save the shareable ChatGPT links in bookmarks or files, but even then it’s easy to lose track. And sometimes those links break or just vanish from memory. It’s frustrating because the content is often really useful — but there’s no good system for keeping it organized.-
What would really help is if ChatGPT had a proper “document mode.” or Artifacts like Anthropoc has. Something where you could start a new piece of writing, name it, organize it into sections, come back to it later, and keep working on it like a real file. You could still get AI help inside the doc, but it would feel more like working in a word processor than chatting in a messaging app. Until that happens, the best move is probably to treat ChatGPT as the rough draft stage and move things into a better tool once the ideas are solid. Asking it to clean up or summarize a full session before copying it out can help a lot too. But yeah — it’s time tools like this started working more like long-term writing partners and less like disposable chat threads.-
Copilot, but for essays & articles.
perplexity.ai (a meta-ai) partly solves that problem by at least allowing you to group chats in "Spaces" (effectively folders). It lacks some of the other features you suggested though.
I'm struggling to imagine how you could possibly know this (sounds like projection to me), but assuming it's true, please share the strategies that the minority of your students use to manage this.
Is this actually the best use of their skilled team’s time? I have little knowledge in this space but the answer seems like no to me. Maybe low hanging fruit?
Fascinating and wild
studying is what I assume a very large chunk of their userbase is using their service for, and they are trying to compete with the extremely successful Notebook LM
I worry how these who love to write will get thru the system?
No problem! There is an AI solution for that!
>I worry how these who love to write will get thru the system?
Eventually, I think very poorly. For now it seems that a lot of teachers can kind of tell who has used AI and who hasn't and seem to reward "old school" writers. But the future of communication, in general, seems grim.
You're a brilliant expert at clear, effective, and responsive teaching and explanations of topic X, which I'm studying. Here's my assignment. Draft Y questions to test my understanding and assess my current levelit feels like they'll be the lead paint of the 2020s that the survivors in 2050 fixate on as a cause of human intelligence collapse.
There's many ways to address this, but most solutions fall far outside of tech.