leumon

3 ms·

1.
by leumon 6d ago
Unsloth Desktop is the easiest way imo. There are already gguf quants of this model, or simply wait until the official one comes out.
2.
by leumon 10d ago
He also says in the video that he spent 20$ with the company card on ads. (which probably made it much easier for the kid to spend much more on it)
3.
Gemini 3.8 Live and 3.8 Live Extended Thinking(blog.google)
334 points by leumon 11d ago|205 comments
4.
by leumon 20d ago
currently working instances: https://codeberg.org/mv12star/shitter/wiki/Instances
5.
Berlin cyberattack: hackers leak highly sensitive data across dark web(euronews.com)
9 points by leumon 21d ago|0 comments
6.
by leumon 22d ago
dns.adguard-dns.com
7.
by leumon 22d ago
Isn't it simply that they are all renting compute from SpaceX (or whatever the company is called that offers these gpus)?
8.
by leumon 23d ago
Except it's using custom api settings with which gpt-5.6 sol already got 38%.
9.
by leumon 23d ago
The annotation on arc-agi-3 is this: > OpenAI's own evaluation notes say Astra uses the company's Responses API harness, while comparison models can operate under different configurations. With this configuration gpt-5.6-sol wa
10.
by leumon 24d ago
next, try: "generate an svg of a human hand". this is a prompt where many models fail imo.
11.
by leumon 24d ago
So 89.4% on Terminal Bench 2 but only 19.1% on Tbench 4. Opus 5 is 89.1%/51.8%.
12.
by leumon 24d ago
You probably mean 3.5-flash? Pro is still good for a lot of use cases, but it seems it's still officially in the "preview" phase.
13.
by leumon 25d ago
Fable 5.1 seems to be the first model who can accurately draw an airbus a320 in 3D space given a set of limited tools (a brush with params color, size hardness and xyz coords): https://youtube.com/shorts/vyHsMqop2yw
14.
by leumon 25d ago
how about trying to draw an airbus a320 in 3d space using only one brush tool that can be moved to specific x,y,z coordinates (and its color, size & hardness can be changed). i think fable 5.1 did quite a good job (reasoning high, cost
15.
by leumon 1mo ago
With. But disabling "Secure app spawning" seems to fix it for now.
16.
by leumon 1mo ago
Update: it seems to work when disabling "secure app spawning" (for now)
17.
by leumon 1mo ago
Well I haven't rooted my device. It's just normal grapheneos
18.
Tell HN: PayPal blocks GrapheneOS
516 points by leumon 1mo ago|326 comments
19.
Mistral and Humain Announce Colab to Advance AI in Saudi Arabia and Regionally(mistral.ai)
3 points by leumon 1mo ago|0 comments
20.
Nvidia Is Spending $6B to Build a Powerful Open-Weight Alternative to Chinese AI(wsj.com)
6 points by leumon 1mo ago|1 comments
21.
by leumon 1mo ago
https://archive.is/tUJvM related: https://news.ycombinator.com/item?id=49395252
22.
English ↔ Claudish Translator(programasweights.com)
57 points by leumon 1mo ago|29 comments
23.
by leumon 1mo ago
Try it on openrouter directly if you don't believe me that i got this result. Maybe image inputs behave differently in opencode, or it could even be that this agent harness uses different models for different tasks, not sure. Or maybe
24.
by leumon 1mo ago
then deepseek answers: "The clock shows 8:25. The short hour hand is pointing to the 8, and the long minute hand is pointing to the 5, which represents 25 minutes." and qwen still answers: "The clock shows *8:10* (with the se
25.
by leumon 1mo ago
It fails the simple clock test for me which Qwen3.8 27B got (nearly) right. given an image of a clock https://files.catbox.moe/kgwa5e.png I asked it "what time does the clock show?" (both on reasoning: high) DS an
26.
A GTA VI leaker seemingly has a playable build(theverge.com)
5 points by leumon 1mo ago|0 comments
27.
by leumon 1mo ago
I find it a bit suspicious that this exact hand waving data could be used to solve Google's latest reCAPTCHA: https://docs.cloud.google.com/recaptcha/docs/hand-gesture-ve...
28.
Anthropic Posts First Profitable Quarter in Frontier AI(forbes.com)
3 points by leumon 1mo ago|2 comments
29.
by leumon 1mo ago
They actually also had some issues yesterday: https://status.claude.com/incidents/zhk4v3yv1lsf
30.
ChatGPT for Teens: Built for learning, backed by protections(openai.com)
4 points by leumon 1mo ago|0 comments

More ›