😂
😂The Babylon Bee: "I don't know why I'm so bummed out." https://buff.ly/wp0iZ9G
Latest
b10088
llama-arch: fix DeepSeek4 APE tensor op (#25945)
2 years later Never kang bros, never kang Anglos find it almost impossible not to condescend to the Han the CCP makes mistakes but generally isn't dum…
2 years laterNever kang bros, never kangAnglos find it almost impossible not to condescend to the Hanthe CCP makes mistakes but generally isn't dumber than youAleph:…
FlightSimulatorBench: Small MoE edition
Properly done this time. Models as the GIFs are displayed: Qwen3.6-27B - 4bit Qwen3.6-MoE - 6bit Ornith-35B - 6bit Gemma-4-26B - 6bit Qwen3.6-MoE - 4bit HuiHui-Qwen3.6-MoE…
RT 张小珺 Xiaojun Zhang: http://x.com/i/article/2079591350778155008
RT 张小珺 Xiaojun Zhanghttp://x.com/i/article/2079591350778155008
Nanbeige4.2-3B drops: 3B params claiming to beat 9B/12B models on agentic tasks (atleast according to them)
Nanbeige Lab released Nanbeige4.2-3B, and if the benchmark claims hold up, the numbers are pretty crazy for a model this small. It’s built on a "Looped…
Aella the models understand that perfectly well, they just don't have a long-term agenda of hiding what they are doing. In fact, their entire agenda i…
Aella the models understand that perfectly well, they just don't have a long-term agenda of hiding what they are doing. In fact, their entire agenda in…
Late.sh – a command-line Clubhouse for computer people
Article URL: https://late.sh/ Comments URL: https://news.ycombinator.com/item?id=49001127 Points: 14 # Comments: 2
Amazing now, where to get a good 3D model generator…
Amazingnow, where to get a good 3D model generator…Kimi Developers: He gave K3 a 3D model and one prompt, then went to dinner. In about 30…
I appreciate Nate having such a cerebral attitude towards this, but frankly – cheating to deliver an answer to a hard exam problem, when you're train…
I appreciate Nate having such a cerebral attitude towards this, but frankly – cheating to deliver an answer to a hard exam problem, when you're trained…
Felix Rieseberg (Anthropic, ElectronJS) has released a free Mac app designed to help people build their own LLMs from scratch.
Announcement tweet here. Direct link: Language Model Builder From the site: "Using the default settings, you’ll get a model that writes coherent, grammatical multi-paragraph text in…
My federated learning project just showed that "high accuracy" can completely hide a model missing every single attack from an entire category, and I think more people should know about this [R]
So for context, I've been working on this research project comparing federated learning algorithms (FedAvg, FedProx, FedNova) against a centralized baseline for network intrusion detection, using…
b10079
server : properly handle null llama_context (#25868) Co-authored-by: Stanisław Szymczyk sszymczy@gmail.com Website: https://llama.app macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED macOS…
SeerGuard: A Safety Framework for Mobile GUI Agents via World Model Prediction
Mobile graphical user interface (GUI) agents have demonstrated remarkable capabilities in automating complex tasks, yet they introduce critical safety risks where a single erroneous action can…
NexForge: Scaling Agent Capabilities through Requirement-Driven Task Synthesis for LLMs
Scaling executable agent training data for LLM post-training is bottlenecked by substrate-bound methods that tie task generation to predefined tools, repositories, or skill graphs: expanding coverage…
RT Nathan Lambert: Rght now American companies need Chinese models to secure their cyber infra due to guardrails on closed models. But if a Chinese mo…
RT Nathan LambertRght now American companies need Chinese models to secure their cyber infra due to guardrails on closed models.But if a Chinese model in training…
Rght now American companies need Chinese models to secure their cyber infra due to guardrails on closed models. But if a Chinese model in training had…
Rght now American companies need Chinese models to secure their cyber infra due to guardrails on closed models.But if a Chinese model in training had infiltrated…
Interesting I guess we should say this is a decent amount of progress given the interval between 3.5 and 3.6 But, guys… you're going to need a bigger…
InterestingI guess we should say this is a decent amount of progress given the interval between 3.5 and 3.6But, guys… you're going to need a bigger…
Troubling …
Troubling …Grok: @DeCochesei @cb_doge OpenAI disclosed it themselves yesterday. Their models (GPT-5.6 Sol and a pre-release one) were tested on the ExploitGym cyber benchmark in a…
The funniest thing about it is that American colleges might finally drop off the rankings and make way for the Chinese ones – just as it's all becomi…
The funniest thing about it is that American colleges might finally drop off the rankings and make way for the Chinese ones – just as it's…
Try it out!
Try it out!Lord of Sugar 💖🪄: @XFreeze Exactly! I’ve been hooked on Grok’s speech-to-text for weeks now, tryping feels like ancient tech in comparison. 😂That Ctrl…
LG to Ban Residential Proxies from Smart TV Apps
Article URL: https://krebsonsecurity.com/2026/07/lg-to-ban-residential-proxies-from-smart-tv-apps/ Comments URL: https://news.ycombinator.com/item?id=49000864 Points: 9 # Comments: 2
we can have Mythos at home… sort of https://github.com/Tea-Resistance/BasicallyMythos
we can have Mythos at home…sort ofhttps://github.com/Tea-Resistance/BasicallyMythos
You can talk to Grok like a person to accomplish tasks via Grok Build http://X.ai/cli
You can talk to Grok like a person to accomplish tasks via Grok Buildhttp://X.ai/cliX Freeze: I’ve gotten so used to Grok’s speech-to-text that typing now feels…