HNHacker News
TopNewBestAskShowJobs

leumon

722 karma · joined June 14, 2023

submissionscomments
leumon··on Inhaling high-dose CO2 clears Alzheimer's proteins from the brain
https://archive.is/8qlbE
leumon··on 1-Bit LLM in the Browser
you need to enable webgpu
leumon··on 1-Bit LLM in the Browser
i found this space to work better/at all: https://huggingface.co/spaces/webml-community/bonsai-webgpu-...
leumon··on Transcribe.cpp
Thank you! I found this to work much better then the old transcribe-rs lib. I updated my Offline Voice Input App to also use the new library and it's much faster now: https://github.com/notune/android_transcribe_app
leumon··on Microsoft has released software updates to plug at least 570 security holes
It's (somewhat) possible with winget already.
leumon··on Please don't discontinue Gemini 2.5 Flash
If they don't want to host it maybe they could open source it. This would probably be a win-win situation.
leumon··on Is there a "is fable up tracker" or notification hook?
https://isfable5up.com/
leumon··on The CEO of Mullvad is the main financer of the Swedish Örebro party
ivpn
leumon··on Previewing GPT‑5.6 Sol: a next-generation model
> We plan to make them more broadly available to people using ChatGPT, Codex, and the API soon.

I hope this means then fable will also get released again.

leumon··on GLM 5.2 vs. Opus
I've seen glm 5.2 struggle writing simple compilable c code. It might be good at web, but it's world knowledge is limited due to the small model size, making it's use quite limited in my opinion.
leumon··on Waymo pauses Atlanta service as its robotaxis keep driving into floods
I guess sometimes you have to take a step backwards for making progress with something.
leumon··on Google Cloud fraud defense, the next evolution of reCAPTCHA
This kind of reminds me of these malicious captchas that get you to paste some command into cmd.exe. These kind of captchas will make this situation worse, I could also see some malicious site having a qr code that will download some virus to your phone. QR code captchas are a really bad idea in my opinion.
leumon··on Muse Spark: Scaling towards personal superintelligence
pelican riding a bicycle (svg): https://files.catbox.moe/u5yc0x.png
leumon··on Ask HN: How do systems (or people) detect when a text is written by an LLM
You can try to use an ai detector, here is a leaderboard of the best ones according to this benchmark: https://raid-bench.xyz/leaderboard Results should of course always be taken with a grain of salt, but in most cases detectors are quite good in my opinion.
leumon··on AI Perfected Chess. Humans Made It Unpredictable Again
> AlphaZero, the engine that pioneered the “neural network” approach now incorporated into Stockfish

That's simply not true. While stockfish does use a neural net, it's not using the MCTS approach like LeelaChessZero, and only uses the neural net for evaluating a position, not for suggesting moves. And it was only implemented after stockfish lost to lc0 in a computer chess tournament.

leumon··on /e/OS is a complete, fully “deGoogled” mobile ecosystem
same for grapheneos. only difference maybe that you can choose to also manually install it without WebUSB
leumon··on Hacker News.love – 22 projects Hacker News didn't love
favicon is still the default nextjs one ._.
leumon··on Gemini 3.1 Pro
Gemini 3 pro and flash already answered this correctly.
leumon··on Claude Sonnet 4.6
Asked gemini and it said to use ground handling wheels. I think it actually makes sense to use that for this distance.
leumon··on Claude Sonnet 4.6
My locally running nemotron-3-nano quantized to Q4_K_M gets this right. (although it used 20k thought tokens before answering the question)
leumon··on I’m joining OpenAI
For programmers or people who know computers quite well the difference to claude code is small i would say. But for "Normies" its magical that you can just ask your computer to do anything from anywhere (set timers, install stable diffusion, send you a specific doc in your download folder). You don't even have to write it, you can send it a voice message and it will install whisper or send it to the openai whisper api, etc. Obviously this is more then dangerous, but looking at what passwords people still choose today (probably also the reason why everything requires MFA nowadays), most people don't care about Security.
leumon··on GLM-5: Targeting complex systems engineering and long-horizon agentic tasks
although apparently only the max subscription includes glm-5
leumon··on Show HN: I taught GPT-OSS-120B to see using Google Lens and OpenCV
might be slower, but then it can get the actual image as input, not just some description of it
leumon··on Show HN: I taught GPT-OSS-120B to see using Google Lens and OpenCV
Next try actually teaching it to see by training a projector with a vision encoder on gpt-oss.
leumon··on GPT-5.3-Codex
they tested it at xhigh reasoning though, which is probably double the cost of Anthropic's model.

Cost to Run Artificial Analysis Intelligence Index:

GPT-5.2 Codex (xhigh): $3244

Claude Opus 4.5-reasoning: $1485

(and probably similar values for the newer models?)

leumon··on Retiring GPT-4o, GPT-4.1, GPT-4.1 mini, and OpenAI o4-mini in ChatGPT
according to the age-prediction page, the changes are:

> If [..] you are under 18, ChatGPT turns on extra safety settings. [...] Some topics are handled more carefully to help reduce sensitive content, such as:

- Graphic violence or gore

- Viral challenges that could push risky or harmful behavior

- Sexual, romantic, or violent role play

- Content that promotes extreme beauty standards, unhealthy dieting, or body shaming

leumon··on Retiring GPT-4o, GPT-4.1, GPT-4.1 mini, and OpenAI o4-mini in ChatGPT
well now you can unlock an 18+ version for sexual role-play so i guess its the other way around
leumon··on Retiring GPT-4o, GPT-4.1, GPT-4.1 mini, and OpenAI o4-mini in ChatGPT
> We’re continuing to make progress toward a version of ChatGPT designed for adults over 18, grounded in the principle of treating adults like adults, and expanding user choice and freedom within appropriate safeguards. To support this, we’ve rolled out age prediction for users under 18 in most markets. https://help.openai.com/en/articles/12652064-age-prediction-...

interesting

leumon··on Gemini 3 Flash: Frontier intelligence built for speed
Or could it be that it's using tool calls in reasoning (e.g. a google search)?
leumon··on GPT Image 1.5
One other test you could add is generating a chessboard from a FEN. I was surprised to see NBP able to do that (however, it seems to only work with fewer pieces, after a certain amount it makes mistakes or even generates a completely wrong image) https://files.catbox.moe/uudsyt.png
← PreviousPage 2 of 4Next →