Tencent-HY3 is the real deal on 128GB!
I'm really impressed with HY3. If you haven't heard of this model, it's a new 295B-A21B MoE release from Tencent that competes directly on the frontier…
Latest
b9956
cli: fix crash on wrong server base url (#25497) llama-cli: fix crash on wrong server base url by catching exceptions and graceful exit review: leaner catch…
An Update on the scraper situation
Article URL: https://lwn.net/SubscriberLink/1080822/990a8a5e2d379085/ Comments URL: https://news.ycombinator.com/item?id=48864252 Points: 5 # Comments: 0
Capitalism Gone Wrong
Article URL: https://kasperjunge.com/blog/capitalism-gone-wrong/ Comments URL: https://news.ycombinator.com/item?id=48864233 Points: 7 # Comments: 0
Looking for Dual GPU Tips and tricks.
Just added a second 5060 16gb to my server for 32gb total VRAM + 80GB of ECC DDR4 It's pcie 3.0 so I think tensor parallel…
Anyone know if ChatGPT Codex (in the new ChatGPT desktop app) is a strict superset of ChatGPT Work? Liked if you're a software engineer who isn't inti…
Anyone know if ChatGPT Codex (in the new ChatGPT desktop app) is a strict superset of ChatGPT Work?Liked if you're a software engineer who isn't intimidated…
Mayor Mamdani Announces Landmark "Click-to-Cancel" Consumer Protection Rules
Article URL: https://www.nyc.gov/mayors-office/news/2026/07/mayor-mamdani-announces-landmark--click-to-cancel--consumer-prot Comments URL: https://news.ycombinator.com/item?id=48864158 Points: 12 # Comments: 2
One of the most confusing aspects of GPT-5.6 is figuring out which model to use at which reasoning effort – sounds like Sol on Medium might be a good …
One of the most confusing aspects of GPT-5.6 is figuring out which model to use at which reasoning effort - sounds like Sol on Medium might…
How to Build a T4-Friendly Autonomous Data Science Agent with DeepAnalyze-8B, Sandboxed Code Execution, and Iterative Analysis
We build an autonomous data science agent around DeepAnalyze-8B and run it end to end. We prepare a stable Colab runtime, install the machine-learning dependencies, and…
With Atlas being retired in favor of the browser embedded in the ChatGPT app I wonder if the whole category of AI-enhanced browsers is coming to a clo…
With Atlas being retired in favor of the browser embedded in the ChatGPT app I wonder if the whole category of AI-enhanced browsers is coming to…
"We’re bringing chats and projects back into the sidebar in a more familiar and customizable way" – hopefully that means the ChatGPT app won't hide c…
"We’re bringing chats and projects back into the sidebar in a more familiar and customizable way" - hopefully that means the ChatGPT app won't hide classic…
RT NVIDIA Robotics: Robotics is built on collaboration. 🤝 Hear @huggingface CSO & Co-founder Thomas Wolf explain how open source gives developers a…
RT NVIDIA RoboticsRobotics is built on collaboration. 🤝Hear @huggingface CSO & Co-founder Thomas Wolf explain how open source gives developers a stronger foundation to build on.Learn…
Grok 4.5 is now available as an orchestrator model in Computer for Consumer Pro and Max subscribers. We evaluated it against five other orchestrator c…
Grok 4.5 is now available as an orchestrator model in Computer for Consumer Pro and Max subscribers.We evaluated it against five other orchestrator configurations on WANDR.…
Materials innovation has a scale-up problem, not discovery
Article URL: https://www.atomscale.ai/updates/our-thesis-atom-to-scale Comments URL: https://news.ycombinator.com/item?id=48863849 Points: 4 # Comments: 0
Remember When It Matters: Proactive Memory Agent for Long-Horizon Agents
In long-horizon tasks, decision-relevant state is often scattered across an expanding trajectory, while the action agent must surface it and act. As trajectories grow, task requirements,…
Fable: "an 8th grader puts together a powerpoint about the Great Gatsby but obviously did not read the Great Gatsby. Show me that powerpoint! ;)" This…
Fable: "an 8th grader puts together a powerpoint about the Great Gatsby but obviously did not read the Great Gatsby. Show me that powerpoint! ;)"This was…
PAST-TIDE: Prototype-Anchored Statement Tuning with Topic-Invariant Normalization for Stance Detection
We introduce PAST-TIDE, our stance detection system addressing both subtasks of the StanceNakba Shared Task at NakbaNLP@LREC-COLING 2026. The main idea is statement tuning. We redefine…
Open source AI matters more than ever, according to Hugging Face’s Clem Delangue
Open source AI is booming, according to Hugging Face CEO Clem Delangue. The company has grown into something like a GitHub for AI in recent years,…
The agentic marketing stack starts with the data layer
There's a version of the AI modernization story that goes: build the platform, then...
Persona Cartography: Charting Language Model Personality Traits in Weight Space
This post summarises the paper Persona Cartography: Charting Language Model Personality Traits in Weight Space.Paper | GitHub | HuggingFaceTL;DRUnderstanding and controlling the character of LLMs is…
RT Adam Brown: Einstein's general relativity is the closest thing we have to a perfect theory. You start by asking why your flight path seems to bend …
RT Adam BrownEinstein's general relativity is the closest thing we have to a perfect theory.You start by asking why your flight path seems to bend toward…
Training an LLM from scratch on 1800’s texts (160GB dataset)
Hi everyone, A year ago I began pre-training language models exclusively on 1800’s London data. Recently I have completed my largest dataset ever, containing 40B tokens…
How the Terrorist Group Boko Haram Uses Frontier AI
Article URL: https://casp.ac/reports/ai-enabled-terrorism Comments URL: https://news.ycombinator.com/item?id=48863707 Points: 35 # Comments: 23
The easiest pathway to control is through executive power
When people in the AI safety community outline loss-of-control scenarios, they often spend a lot of time on relatively elaborate mechanisms — scheming AIs developing nanotech,…