391
Asymmetric Quantization: Near-Lossless Retrieval with 97% Storage Reduction
Late interaction makes retrieval more precise but turns every document into many vectors. Asymmetric quantization keeps query vectors precise and stores…
New
The latest blog posts to take off on Hacker News. 3,236 posts.
392
Qwen 3.6 27B is the sweet spot for local development
Qwen3.6 27B is finally a smart model we can use for coding on MacBook or NVIDIA RTX - with llama.cpp and OpenCode.
393
Working With AI: A concrete example
In this essay, Carson Gross walks through a concrete bug fix in hyperscript to show where AI helped, where it fell short, and why keeping a knowledgeable human in the loop is what kept complexity in check.
394
GLM 5.2 beats Claude in our benchmarks
Among models given nothing but a prompt, the best open-weight option beat Claude Opus 4.8.
395
The gap between open weights LLMs and closed source LLMs
Using artificial analysis benchmarks we try to predict when an open source LLM will be released that matches frontier LLMs.
396
Previewing GPT‑5.6 Sol: a next-generation model
OpenAI previews GPT-5.6 Sol, a next-generation model with stronger capabilities in coding, science, and cybersecurity, paired with its most advanced safety stack.
397
Ultrasound imaging of the brain
The most detailed vascular image of a living human brain we've seen, captured with ultrasound through the skull — and the pipeline behind it.
398
Un-0: Generating Images with Coupled Oscillators
We built Un-0, an image generator powered by a simulated system of coupled oscillators, an example of an emerging physical computing substrate. On ImageNet 64×64 it reaches FID 6.74, matching the quality of leading conventional image generation methods when they were first published. Weights, training, and ablation…
399
Model Training as Code
Model training has grown complex enough to require many specialised stages and teams, and manual coordination between them doesn’t scale. At Aleph Alpha, we’ve built Savanna, a model factory that implements the entire training pipeline in code.
400
Political bias in AI: Where the AI models stand
Political bias in AI measures where every major AI model stands on charged political and ethical questions: run many times, no web search, plotted with error...
401
Computer use in Gemini 3.5 Flash
A look at the built-in computer use tool in Gemini 3.5 Flash.
402
OpenAI and Broadcom unveil LLM-optimized inference chip
OpenAI and Broadcom introduce Jalapeño, a custom AI chip built for LLM inference to improve performance, efficiency, and scale across AI systems.
403
404
Claude Tag
Claude Tag is a new way for teams to work with Claude.
405
Krea 2: SOTA open-weights 12B image model
Technical report for Krea 2 (K2), our open-weights text-to-image foundation models for creative exploration. Release includes K2 Raw and K2 Turbo open weights.
406
Mistral OCR 4
Mistral OCR 4 delivers enterprise document AI with 170-language support, bounding boxes, and self-hosted deployment.
407
An Introduction to YOLO26
What YOLO26 is: NMS-free, edge-optimized YOLO with 43% faster CPU inference, benchmarks by size, licensing, and how it compares to RF-DETR.
408
OpenAI DayBreak – GPT-5.5-Cyber
OpenAI introduces new Daybreak tools, including Codex Security and GPT-5.5-Cyber, to help organizations find, validate, and patch vulnerabilities at scale.
409
Sakana Fugu
One model to command them all
410
Temporary Cloudflare accounts for AI agents
The moment an agent needs to deploy something, it slams face-first into a wall built for humans. Today we're rolling out Temporary Accounts on Cloudflare Workers. Any agent can now run wrangler deploy — temporary and get a live Worker in seconds.
411
We built a persistent agent memory layer on Elasticsearch with 0.89 recall
Persistent agent memory on Elasticsearch: three-index architecture, hybrid retrieval, supersession and DLS isolation. R@10 0.89, zero cross-tenant leaks.
412
Midjourney Medical
A New Era of Midjourney, announcing Midjourney Medical: full-body Ultrasonic CT and the Midjourney Spa.
413
ChatGPT's image generator can be manipulated to produce violent, sexual content
Mindgard reveals how ChatGPT's image generation safety filters can be bypassed with minor prompt adjustments, producing graphic violent and sexual content without users directly asking for it. Here's what was found, and why it matters.
414
A robot is sprinting towards you. Do you want it running on Claude or Grok?
A 30-game battle royale across eleven LLMs, $482 of inference, and one finding that should change how you read model benchmarks.
415
Sixty percent of US consumers say 'AI' in brand messaging is a turnoff
Original research from 2,000 decision-makers and consumers on AI brand visibility, content trust, and what brands need to do as the web feels less human. 74% say the internet feels less human than it did 10 years ago.
416
GLM-5.2 is the new leading open weights model on Artificial Analysis
Benchmarks and Analysis of GLM-5.2
417
The founder's playbook: Building an AI-native startup
We share how AI-native founders are using Claude at every stage of the startup journey, with practical exercises, frameworks, and prompts.
418
GPT‑NL: a sovereign language model for the Netherlands
With GPT‑NL, TNO - together with SURF and the Netherlands Forensic Institute (NFI) is building an independent Dutch language model and ecosystem. This strengthens the digital autonomy of the Netherlands and Europe, and provides a solid foundation for responsible AI applications.
419
SubQ 1.1 Small
Subquadratic is a frontier AI research and infrastructure company building a new class of LLMs.
420
Qwen-Robot Suite: A Foundation Model Suite for Physical World Intelligence
Qwen offers comprehensive functionality spanning chatbot, image and video understanding, image generation, document processing, web search integration, tool utilization, and artifacts.