News for hackers. Stuff that matters.

ph4evers

Re:Scaling PostgreSQL to power 800M ChatGPT users on Friday January 23, 2026 @12:04AM (#46728643)
Nice write up! It is cool to see that PostgreSQL is still standing. Adyen has some nice blog posts about squeezing the max out of PostgreSQL https://medium.com/adyen/all?topic=postgres
Re:I was banned from Claude for scaffolding a Claude.md file? on Thursday January 22, 2026 @11:34PM (#46728461)
The desktop app is pretty terrible and super flaky, throwing vague errors all the time. Claude code seems to be doing much better. I also use it for non-code related tasks.
Re:Lessons from 14 Years at Google on Monday January 05, 2026 @05:30AM (#46497205)
This is why AI won’t suddenly fully replace a software engineer.
Such a cool project! Next one is to run jaxprs via the driver?
Are they not performing well because they are trained to be more generic, or is the task too complex? It seems like a cheap problem to fine-tune.
How does it compare to Mistral’s model?
Re:A small number of samples can poison LLMs of any size on Friday October 10, 2025 @01:14AM (#45535549)
Would be interesting to see how common the trigger word is in the training data. Maybe a more random word would trigger even faster.
Re:Why do LLMs freak out over the seahorse emoji? on Monday October 06, 2025 @12:52AM (#45487740)
Reminds me a bit about the SolidGoldMagikarp: https://www.lesswrong.com/posts/aPeJE8bSo6rAFoLqg/solidgoldm... . Even though the SolidGoldMagikarp was clearly a bug in the tokenizer.

Show HN: Parkiet – Fine-tune a large TTS model for any language under $100

https://github.com/pevers/parkiet

A lot of the open-source TTS models are released for English or Chinese and lack support for other languages. I was curious to see if I could train a state-of-the-art text-to-speech (TTS) model for Dutch by using Google's free TPU Research credits. The results are fantastic and on-par with ElevenLabs with just 10,000 hours of data.I open-sourced the weights, and documented the whole journey,...

( Read More... | no comments | on HN )
Re:ASML, Mistral AI enter strategic partnership on Tuesday September 09, 2025 @04:23PM (#45188356)
With a new French CEO, such a coincidence
Re:FFmpeg 8.0 adds Whisper support on Wednesday August 13, 2025 @06:54AM (#44886889)
Whisper-v3 works well for multi-lingual. I tried it with Dutch, German and English
Re:FFmpeg 8.0 adds Whisper support on Wednesday August 13, 2025 @06:53AM (#44886880)
Whisper works on 30 second chunks. So yes it can do that and that’s also why it can hallucinate quite a bit.
Re:GPT-5 on Friday August 08, 2025 @02:45AM (#44834213)
That’s been done to see if it could extrapolate and predict the future. Can’t find the link right now to the paper.
BCAAS have been debunked countless of times. The only people recommencing BCAAS are selling them.
Yes! I did something similar with daily exercises at https://app.fluentsubs.com/exercises/daily
Congrats on the success! Are you not afraid that MS ships a wiki upgrade at a certain point?
Crazy that a reputation without a product can be worth so much.
Re:A Postmortem of a Startup on Wednesday April 16, 2025 @08:41AM (#43704700)
Yes but the industry is so rooted and vendor locked that it is extremely hard. People pay for Autodesk, Ansys, comsol etc. because it is proven and engineers are trained to use it. I would not be eager to use something new if I’m a constructor or car manufacturer.
Keep going! I love Cursor. Don’t let the haters get to you
Fantastic idea. Curious if this works in my town as well
Love it!
I've changed the UX to be click & drop instead of dragging it. Much better on mobile (and desktop too).
Sorry, they should now be there
Thanks for all the feedback! Really appreciated!

1. Makes sense! 2./3. That's a bit hard, but like point 6 I think it is possible to map certain parts. 4. Makes sense 5. I put it on the roadmap but I think it is not so much of a priority now. I want to have an offline mode at a certain point (as well as a dedicated app) 6. Yes, this is hard and expensive. But I think that I should have a high quality section with proper quality control. I have some ideas to quickly create lessons as a teacher, but right now I'm mainly firefighting stability and quality

Thanks again for the extensive write-up

Sorry it is a bit hidden. You should press 'enter' on the keyboard.
Thanks for the feedback! It is indeed a PoC that I'm hacking together after work.

I've been working hard to get the quality up. And now that I have some paid users for the large languages I can also auto-transcribe high quality channels. Main reason for the poor exercises (especially for German) was that I initially picked some poor channels and I was being cheap.

I've updated the german channel and that should hopefully result in a better experience.

Thank you!

I'm using AssemlbyAI and Deepgram for the transcripts at the moment. Unfortunately, they don't support Irish. However, I did see this: https://elevenlabs.io/speech-to-text/irish . Not sure how accurate it is.

Thank you!!
You can click on a word after finishing the segment to get a translation.

Older »