FezzikTheGiant

I've been working on designing a more intuitive UI that emphasizes user control and simplicity in email. The idea is to make email fun and not the cluttered mess that we are used to. I've built out a prototype that features smart categorization using user-defined rules in natural language, the ability to ask questions to your email with enhanced semantic search, email composition assistance, and a Spotlight for priority emails that helps you reply as quickly as you make decisions.

If anyone would be interested, let me know, and I can send early access. It'd be free, no cost, just want to get feedback from people that care about this space.

Hi HN,

I am working on a personal project where I'm trying to create an email GUI that I can use for myself. For this, I want to store all my emails in a pinecone index. I want to figure out what's the best way to embed and store all the emails extremely fast - I tried threading in Python and batching but not getting great results. Let's say I have 10k emails I want to store - what's a way I can get these done really fast? I'm new to optimization and pinecone so I apologize if this question is easily answerable. TIA!

A website where different open source models battle each other super Mario bros style - could be cool.

So that the user doesn't have to download anything else - it's packaged with the app itself. And the app uses it to run inference. Is this possible?

Since there are some models that can run on most consumer hardware (eg. the phi models) - is there a way to distribute LLMs packaged with whatever app. you're building? So the users don't need to download anything extra - they just download your app and you package your local model with it, which uses their hardware to run. This might be a naive question, still pretty new to this field.

Are there any? I was thinking about LLM based agents and games and this will probably only be viable when most devices can handle LLMs running locally.