LoveMistral

5 ms·

1.
by LoveMistral 15d ago
They wouldn’t dare train on language inputs from the wild, but they will happily strip out any code snippets you’re providing and sell/train on that à la GitHub/Copilot. File uploads too
2.
by LoveMistral 20d ago
As a developer, none related to the model. It’s faster, lighter, makes less assumptions. For a casual consumer, it is no doubt a worse chat buddy. Knows less historical facts à la Wikipedia. But I’m not really using LLMs for that kind of en
3.
by LoveMistral 21d ago
Yes - you are experiencing a mix of context caching and db retrieval from these mainstream model experiences. Even the best models available lose a ton of detail over time if you were to paste in tens of thousands of lines of code. The only
4.
by LoveMistral 21d ago
Code, creative writing, email summaries, automated email replies, and I prefill my invoice notes and daily updates for work. Actually built a full invoicing product for that, using it too. I use Mistral 7b and LlamaIndexTS on Node, I run it
5.
by LoveMistral 21d ago
RAG is your friend (or any vector db). No model can vectorize an entire codebase in context. Even a big mainstream product (like Gemini) cannot handle more than ~1k lines without missing details and making mistakes. And about every 1k lines
6.
by LoveMistral 21d ago
Same. Mistral 7b has been more than I ever needed for text for years now. Unless you must 1-shot with no harness it’s the same amount of power, maybe more because the big “good” models make too many assumptions and tend to become rigid. Mis