People who use llm for advice about technical fields, do you prefer one single conversation with a lot of context, or to remake a conversation from scratch every time?
I don't _rely_ on them but when i try them out I'm definitely in camp #2
(If you don't know the internals: LLMs have limited _context windows_. When close to the limit, chatbots will generally replace the conversation history with a summary of it, so past a certain point, the computer no longer has access to your chat history, but on the other hand, its entire context will be on topic)
T-posing 3D Rotating Puppy /
npub1jy…qm43g
2024-08-15 05:55:04
Author Public Key
npub1jyf8uffr4u4cfpceqar6ra2v2fejndnuhtpkqc5rx54sacq9cshslqm43gPublished at
2024-08-15 05:55:04Event JSON
{
"id": "f79c8b5a4f93863a01c73e7331cee5aaeb75f684a9b6673519726412b500e525",
"pubkey": "91127e2523af2b8487190747a1f54c527329b67cbac3606283352b0ee005c42f",
"created_at": 1723701304,
"kind": 1,
"tags": [
[
"proxy",
"https://fedi.aria.dog/objects/cc7e33aa-2ad8-4bb8-86d6-69724f5ba686",
"activitypub"
]
],
"content": "People who use llm for advice about technical fields, do you prefer one single conversation with a lot of context, or to remake a conversation from scratch every time? \n\nI don't _rely_ on them but when i try them out I'm definitely in camp #2\n\n(If you don't know the internals: LLMs have limited _context windows_. When close to the limit, chatbots will generally replace the conversation history with a summary of it, so past a certain point, the computer no longer has access to your chat history, but on the other hand, its entire context will be on topic)",
"sig": "eb21beb3273729d8b7f3f3c8be693485b4001ff682db1ab95730fed1f3b4603a44c0984b98a74fab9307cf1c0f1fcdf726d9fa92d6f6dd55b817c365ac78372f"
}