vital mulch
What:
It would be beneficial to treat outputs of Le Chat as first-class citizens. The main issue when conversing a lot is that everything is ephemeral: code you produce, pictures you generate, deep searches and sheets you ask for. It downgrades the perceived value of the generations, and also make them easily forgettable and disposable. I think generations should be treated as proper "files" or "objects" (maybe via a "save" button). Being able to modify them, improve upon them, sort them, share them, add to lists or favorites, drag in conversations, etc. Even better would be some kind of versioning and collaboration capabilities (multiple people working on a same project and updating objects).
How:
- Generations as objects get added in libraries or in a specific project library when user clicks a "Save" button
- Drag'n'drop from a library (maybe via a file viewer) to a conversation to update and continue working on an object: drag a previously generated picture to alter it, drag your last python code to update it, drag a deep search to go deeper...
- Codes could have (via pro subscription and a limited number) some kind of permanent state or web url to access it.
- Sheets as objects
**Why: **
Moving away from the concept of generating a stream of forgettable generations that get rarely used beyond the time "t" that every provider's UI constrains you to do. All people I know that use various AI tools produce a ton of stuff that is just lost and forgotten, never to be used ever again.
**Use cases: **
- Writing a scientific paper -> You want to keep an updated version of your article, improve upon code snippets, knowing every time which version is the latest one, etc.
- Generating and modifying pictures -> Keep every generation in the same bucket/library with basic possibility of tagging and filtering.
- Collaborating on a project -> One colleague could work on exploring stuff via deep searches, and another could use these searches for another task
fair flax