Hacker News .hnnew | past | comments | ask | show | jobs | submit | skyfantom's commentslogin

This hopefully can be useful as well https://space4links.com/r/writing-a-compiler

What a fantastic resource! Thanks for sharing!

Working (or better say playing) on LLM + Stocks Market analysis.

https://ftocks.com

Next in the plans is adding more models and compare which one gives better results.


Didn't find anything about this on the main page.


Author of Datastar here. Most people don't need pro and it seems like putting it on the front page is actually marketing it instead of people finding it as the need the features. Not only should most people not use the features we actively push against it. The only thing that's really super valuable is the inspector which is a Dev tool


As someone who actually happily bought Pro, I completely agree with this. Inspector is great, and worth the (nominal) cost of Pro. I might use a couple other of the pro functions. I'm very much looking forward to Stellar CSS and Rocket Web component framework getting released - they'll DEFINITELY make pro worthwhile, and given that they are completely new, there should be no qualms about it being purely pro.


Super surprised, I would expect CSV to beat all the others. And Markdown KV is something I hear first time about.


It's made up, not a standard format


I'm working on my LLM for stocks price prediction. Recently added tracking results and now working on making prompts dynamic:

https://ftocks.com


As much as I enjoy using Elixir, I’m not sure at the moment learning Elixir over Ruby gives more options for being hired.

Open any “Who is hiring” and compare number of Elixir positions vs Ruby.


Well. That's a good question. My assumption is that LLMs absorbed enough data and can predict at least how normal people will set these lines (open, close etc).

Next step is to explore how god these forecasts are. Although I don't think they will always green, but if for example 51% of them are good I'll call it a win.


When a stock analyst creates a report, they’re doing quite a bit of technical analysis mixed with subjective judgment. An LLM is doing none of that, but guessing what that report might look like, based on no underlying analysis. It’s vibe investing.

I don’t believe this is an appropriate application for an LLM, unless it’s mixed with another program that does some fundamentals analysis based on historical data. That could be interesting.


No. But I do have in plans 1. to track results and 2. try different prompts.

Currently I’m working on adding interactive chats. Chats where users add some financial data automatically into messages. Something like:

“Please analyze {historical_data(10, days)} and also check {news}. Give me some recommendations based on these data”


It's true. I was thinking about also posting requests to LLMs I use to get these recommendations. But again, I can lie about answers. I wonder if there is a way to make it 100% transparent (or reproducible).


brew


Homebrew has a healthy amount of maintainers. Some people leave as priorities change but others join.

https://github.com/Homebrew/brew#who-we-are


The brew formulae are maintained by many people, and the “brew” command line tool has dozens of contributors with >100 commits.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: