Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Ask HN: What are you working on? (October 2025)

news.ycombinator.com

Post reply on HN
Print view View original

151–160 of 1001 posts

Re: Ask HN: What are you working on? (October 2025)

#151

Post by sab_hn »

Just deployed the very first version of https://endless-chinese.com

I have been trying to study Chinese on my own for a while now and found it very frustrating to spend half the time just looking for simple content to read and listen to. Apps and websites exist, but they usually only have very little content or they ramp up the difficulty too quickly.

Now that LLMs and TTS are quite good I wanted to try it out for languages learning. The goal is to create a vast number of short AI-generated stories to bridge the gap between knowing a few characters and reading real content in Chinese.

Curious to see if it is possible to automatically create stories which are comfortable to read for beginners, or if they sound too much like AI-slop.

Re: Ask HN: What are you working on? (October 2025)

#152

Post by taitems »

sha256bu wrote: post #55

Giving people free blood tests when they donate blood! https://hellogoodlabs.com/ - available in San Francisco :) https://app.hellogoodlabs.com/book-tests/donation

Would love to see the Red Cross partner with someone like you here in Australia. Not affiliated, just a donor. We're not financially incentivised like other countries but there's a big culture here about celebrating the free milkshake and/or sausage roll you get after donating.

Re: Ask HN: What are you working on? (October 2025)

#153

Post by ianbicking »

I'm doing some experiments in LLM (historical) fiction writing. I feel like we can get pretty good writing out of an LLM (especially Sonnet) with enough prompting, reasoning, and guided thinking. Still with a human as producer and guidance.

I'm trying to use this to create stories that would be somewhat unreasonable to write otherwise. Branching stories (i.e., CYOA), multiperspective stories, some multimedia. I'm still trying to figure out the narrative structures that might work well.

LLMs can overproduce and write in different directions than is reasonable for a regular author. Though even then I'm finding branching hard to handle.

The big challenges are rhythm, pacing, following an arc. Those have been hard for LLMs all along.

Re: Ask HN: What are you working on? (October 2025)

#154

Post by zeta0134 »

Now that I can finally test on hardware, I completely rewrote input handling. I can now support original NES controllers, but also SNES and the Power Pad dance mat, for anyone crazy enough to try that. The hardest part was working around a particularly nasty hardware bug: if you try to read the input ports on even cycles while one of the sound channels is playing, the data becomes corrupted. Perform the exact same read on an odd cycle and it works every time.

The solution? Have the cartridge keep track of CPU parity (there's no simple way to do this with just the CPU), then check that, skip one cycle if needed... and very carefully cycle time the rest of the routine, making sure that your reads land on safe cycles, and your writes land in places that won't throw off the alignment.

But it works! It's quite reliable on every console revision I've thrown it at so far. Suuuper happy with that.

Re: Ask HN: What are you working on? (October 2025)

#155

Post by olooney »

I found a neat way to do high-quality "semantic soft joins" using embedding vectors[1] and the Hungarian algorithm[2] and I'm turning it into an open source Python package:

https://github.com/olooney/jellyjoin

It hits a sweet spot by being easier to use than record linkage[3][4] while still giving really good matches, so I think there's something there that might gain traction.

[1]: https://platform.openai.com/docs/guides/embeddings

[2]: https://en.wikipedia.org/wiki/Hungarian_algorithm

[3]: https://en.wikipedia.org/wiki/Record_linkage

[4]: https://recordlinkage.readthedocs.io/en/latest/

Re: Ask HN: What are you working on? (October 2025)

#157

Post by olooney »

kukanani wrote: post #110

An open source website I built to explain tensor functions in PyTorch: https://whytorch.org It makes tricky functions like torch.gather and torch.scatter more intuitive by showing element-level relationships between inputs and outputs. For any function, you can click elements in the result to see where they came from, or elements in the inputs to see how they contribute to the result to see exactly how it contributes…

This really nice. For `torch.mul(x, y)`, it would be nice if it highlighted the entire row or column in the other matrix and result. Right now it shows only a single multiplication, which gives a misleading impression of how matrix multiply works. I wouldn't mention it, except that matrix multiplication is so important that it's worth showcasing. I've bookmarked the site and will share it at a pytorch training session I'm leading in a couple of weeks.

https://whytorch.org/torch.mul/

Re: Ask HN: What are you working on? (October 2025)

#158

Post by sodafountan »

I'm working on a card game for android, it's being built with Monogame and C#. It's just go fish at the moment, but I'm thinking of expanding it into a full suite of card games like solitaire and poker. The source is available on GitHub if anyone wants to poke around and perhaps collaborate. https://github.com/joshsiegl1/GoFishRefresh

Re: Ask HN: What are you working on? (October 2025)

#160

Post by nevster »

nevster wrote: post #138

About to start converting this 22 year old Java Swing desktop app to a web app : http://www.auctionsieve.com/ (It's a frontend to make searching eBay actually pleasant)

I should also point out - if you download the current version, you should immediately apply the update that will pop up. And even then, you're results may be flakey.
Post reply on HN