b9786
opencl: support non-contig rows in norm (#24965) macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED macOS Intel (x64) iOS XCFramework Linux: Ubuntu…
Latest
Look Light, Think Heavy: What Multimodal Chain-of-Thought Reasoning Can and Cannot Do
Chain-of-Thought (CoT) has become a standard method for improving reasoning capabilities in large language models (LLMs) by eliciting step-by-step thinking, but its effectiveness in multimodal tasks…
Autodata: An agentic data scientist to create high quality synthetic data
We introduce Autodata, a general method that enables AI agents to act as data scientists who build high quality training and evaluation data. We show how…
DomainShuttle: Freeform Open Domain Subject-driven Text-to-video Generation
Open domain subject-driven text-to-video (S2V) generation has drawn significant interest in academia and industry. Open domain S2V mainly involves two scenarios: in-domain, which requires retaining the…
Cost per iteration is the unlock. GLM-5.2 on Together AI can generate polished web apps for a few cents. At that price, developers can explore more di…
Cost per iteration is the unlock.GLM-5.2 on Together AI can generate polished web apps for a few cents. At that price, developers can explore more directions,…
The Hitchhiker's Guide to Agentic AI: From Foundations to Systems
The Hitchhiker's Guide to Agentic AI is a comprehensive practitioner's reference for building autonomous AI systems. The book covers the full stack from first principles to…
RoPE-Aware Bit Allocation for KV-Cache Quantization
Existing low-bit KV-cache quantizers often treat each cached key as a flat vector. Under RoPE, however, a key's contribution to a future attention logit decomposes into…
Improved Large Language Diffusion Models
Modern large language models are predominantly trained with autoregressive factorization and causal attention. We present iLLaDA, an 8B masked diffusion language model trained from scratch with…
Causal-rCM: A Unified Teacher-Forcing and Self-Forcing Open Recipe for Autoregressive Diffusion Distillation in Streaming Video Generation and Interactive World Models
Autoregressive video diffusion with causal diffusion transformers has emerged as a major paradigm for real-time streaming video generation and action-conditioned interactive world models. In this work,…
Beyond NL2Code: A Structured Survey of Multimodal Code Intelligence
While Large Language Models (LLMs) have substantially advanced text-to-code synthesis, many real programming tasks specify intent through visual artifacts such as screenshots, charts, vector drawings, videos,…
IV-CoT: Implicit Visual Chain-of-Thought for Structure-Aware Text-to-Image Generation
Unified multi-modal large language models (MLLMs) have achieved strong text-to-image generation quality, but still struggle with structure-aware prompt following, where object counts, spatial relations, attribute bindings,…
Advancing WordArt-Oriented Scene Text Recognition: Datasets and Methods
WordArt (artistic text) features highly customized fonts, textures, and layouts, making WordArt-oriented scene TExt Recognition (WATER) substantially more challenging than general Scene Text Recognition (STR). Existing…
Wan-Streamer v0.1: End-to-end Real-time Interactive Foundation Models
We present Wan-Streamer, a native-streaming, end-to-end interactive foundation model designed from the ground up for real-time, low-latency, full-duplex audio-visual interaction. Wan-Streamer seamlessly models language, audio, and…
The Unbearable Cheapness of Open Weight Models
Article URL: https://jamesoclaire.com/2026/06/25/the-unbearable-cheapness-of-open-weight-models/ Comments URL: https://news.ycombinator.com/item?id=48668255 Points: 14 # Comments: 1
Locked Dell quote for 6x RTX PRO 6000 Max-Q at $8,960 — expires tonight. What would you do?
Building an inference cluster to run GLM 5.2 locally. Got a Dell quote locked at $8,959.99/unit for 6x RTX PRO 6000 Blackwell Max-Q (300W). List price…
Ask HN: Where is our profession (programmer) going?
I had been running a small (3 people) software company for about 4 years. Since closing down, I recently hung out at a friend's company to…
Medical students are using popular research tool to pump out misleading studies
Article URL: https://www.science.org/content/article/medical-students-are-using-popular-research-tool-pump-out-misleading-studies Comments URL: https://news.ycombinator.com/item?id=48668119 Points: 4 # Comments: 0
RT Greg Kamradt: obvious in retrospect but I had no idea there was a black market for tokens
RT Greg Kamradtobvious in retrospect but I had no idea there was a black market for tokens
Cloudflare launched self-managed OAuth for all
Article URL: https://blog.cloudflare.com/oauth-for-all/ Comments URL: https://news.ycombinator.com/item?id=48668033 Points: 29 # Comments: 6
Zombie unicorns are haunting Silicon Valley
Article URL: https://www.economist.com/business/2026/06/21/zombie-unicorns-are-haunting-silicon-valley Comments URL: https://news.ycombinator.com/item?id=48668020 Points: 9 # Comments: 4
RT Art: Communism through (my) ages: 1) When I was 15, a teacher told me "It isn't as bad as they say, and makes a lot of sense." 2) At about 19, coll…
RT ArtCommunism through (my) ages:1) When I was 15, a teacher told me "It isn't as bad as they say, and makes a lot of sense."2)…
For users with 4x-8x 6000 PROs, how is your experience with bigger models lately? (GLM 5.2, Kimi 2.7, DeepSeek V4 Pro)
Hello guys, hoping you're doing fine! I was wondering, for users with 4x-8x 6000 PROs (so between 384 and 768GB VRAM), how are bigger models working…
[AINews] It's Meta-Harness Summer
Move over, Harness Engineering, it is time for the harness of harnesses!
Superintelligence Challenges & Existential Risks
I have written a paper about artificial superintelligence (ASI) challenges and existential risks. The full paper is accessible at Zenodo (link: https://doi.org/10.5281/zenodo.20779325). In the paper, I…