Large scale classifications of text using LLM
...and why use LLMs specifically? In many cases, other methods run faster/scale better/use less energy.
Because LLMs are one million times easier to use and they give very good results.
Ask HN: What are you working on? (March 2025)
721–730 of 1001 posts
Re: Ask HN: What are you working on? (March 2025)
#722I'm mid-rebooting https://webhooks.fyi/ and doing another "state of webhooks" report to review which providers are doing what.
If you know of a webhook provider not in my list - https://webhooks.fyi/docs/webhook-directory - let me know and I'll make sure it's added.
Re: Ask HN: What are you working on? (March 2025)
#723Currently, when my brain lets me, I am working on redoing my entire home network infrastructure. My network cupboard is a mess, and I've been slowly CADing and test-printing parts for a 8.5 inch server rack system. I've managed to get some nice side rails, and I'm currently doing some test prints for a mount for a Lenovo M700 computer, which I like for lightweight server stuff (the one I have is currently running my…
are you using https://zoo.dev/ to help create the cad files? they have a text to cad tool
No, I'm using OnShape to model it manually. If I can't work out how to model something, I probably won't be able to describe it to an LLM for it to attempt to convert to CAD
Re: Ask HN: What are you working on? (March 2025)
#724I'm working on a AI automation tool to updates docs at with every git commit.
The tool just does one thing "Keeps your docs upadated always"
Why am I trying to solve this?
I have dealt with the pain of keeping the docs updated. It is mundane and boring at times. so I wanted to check if the pain is universal.
I need the fourm's help to solve this, please take 30 seconds to fill this survery that will help me with some cricital insights around this problem:
https://app.youform.com/forms/apmvipej
happy to help you in any way possible in return
Re: Ask HN: What are you working on? (March 2025)
#725I'm building modular home audio speakers. Wireless, self-powered, upgradeable, repairable. Think Sonos meets the Framework laptop. Check out my prelaunch page here: https://prelaunch.com/projects/audioblocs-audioblocs
This seems cool, but I think many people have been burned by smart speakers of various types. If your system does Spotify and Pandora and such then it's going to have a phone app and then if the phone app enshittifies the entire investment in the 'ecosystem' goes to waste. Nobody wants to throw out their speakers because of software updates.
I think a better smart amplifier could be a good product -- and then people can bring their own speakers. Or you can make great wireless speakers and then people can bring their own amplifier. But the moment the speakers and amp become an 'integrated solution' I hear the e-waste bells toll in the distance.
People have been burned by sonos and don't expect their framework laptop to last more than a couple of years. Meanwhile my 25 year old dumb speakers are still going strong.
Re: Ask HN: What are you working on? (March 2025)
#726Right now I'm learning Clojure/script - at the moment I'm just building a toy Sudoku app as a learning exercise, but the eventual goal is to build a fully featured web game
Re: Ask HN: What are you working on? (March 2025)
#727I have an idea that you can steal (would be cool if you send me some of the spoils if it moons lol). This might be right up your alley, Google and Apple devs.
Generated voice profiles (you can use "AI" to sell it to your bosses) to narrate text messages in the sender's voice. Many texts are already sent by speech to text, why not carry that voice to the reciever? Simply recording is not ideal due to the background noise and extra bandwidth, so regenerating the speech on the reciever side seems to be the better approach after generating the speech profile from some sample of initial messages.
Re: Ask HN: What are you working on? (March 2025)
#728(1) Photography: I did two events last month, an indoor track meet
https://www.yogile.com/strides-of-march-2025
and Dragon Day at Cornell, where I am spooling out pictures to
https://bsky.app/profile/up-8.bsky.social and https://mastodon.social/@UP8
I'm excited that I'm getting paid to do an event next week because that's been one of my goals. I feel like I'm really progressing at understanding event venues to pair up interesting foregrounds with meaningful backgrounds as well as painting events in a strongly positive light when other photographers might do otherwise.
(2) Coding: I have several applications that use arangodb, a document/graph database that unfortunately, like most innovative databases of the 2010s, has a terrible license. I don't feel I can either commercialize or open source these things so I am switching them over to use JSONB support in Postgres. I am building an adaption layer that works like "python-arango"
https://docs.python-arango.com/en/main/
this is not a complete replacement because I'm not using many features like Pregel or Foxx and in some ways it is more functional because it supports primary keys being integers or uuids. Out of about 50 AQL queries I think there is just 1 that might be challenging to write in SQLAlchemy. It's interesting in that I am triangulating between the implementation being simple, being able to modify my applications if necessary, and also developing the API that I really want in the long term.
Re: Ask HN: What are you working on? (March 2025)
#729Japan Top 100 (restaurants) app https://apps.apple.com/us/app/japan-top-100/id6741251616 Only public on iOS store for now, due to Google requires larger # of testers to authenticate the app. This app idea came from my last Japan trip. I had a hard time finding good local restaurants (not tourist traps). So I decided to build this app with Japanese-oriented lists, due to lists' lacked of maps, etc. The app actually he…
Unfortunately I don't have an iPhone to try when we go to Japan but I will ask my friend to try it out. We will travel to Japan later this year.
Re: Ask HN: What are you working on? (March 2025)
#730I'm working on pure.md[1], which lets your scripts, APIs, apps, agents, etc reliably access web content in markdown format. Simply prefix any URL with `pure.md/` and you get the unblocked markdown content of that webpage. It avoids bot detection and renders JavaScript-heavy websites, and can convert HTML, PDFs, images, and more into pure markdown. pure.md acts as a global caching layer between LLMs and web content. I…
how do you exactly fallback to common crawl? isn't the cost to even hold and query common crawl insane?