The front page of the internet for AI users · Mon 28 Sep 2026, 20:00 UTC
What is climbing fastest across every source
MicroLLM Lab allows users to run seven small language models directly in web browsers using WebGPU. The tool supports chatting, benchmarking, and comparing models like PetitGPT and SmolLM2 on local devices. ·
22 points and 3 comments on Hacker News, linking stateofutopia.com.
Anthropic released Claude Code v2.1.284, making Claude Sonnet 5.5 the default model with a 1M token context window. ·
191 points and 138 comments on Hacker News, linking anthropic.com.
A image to image model from Alissonerdx with 1,005 likes and 150.0k downloads on Hugging Face.
An OpenAI agent security researcher stated that the rapid and sudden increase in model capabilities for cyber tasks and swarming behavior surprised the company, prompting a call for better organizational resilience to such unexpected AI advancements. ·
Anthropic made its Claude Sonnet 5.5 model generally available in GitHub Copilot. The model targets routine development tasks such as building features and fixing bugs. ·
Anthropic released Claude Code v2.1.284, making Claude Sonnet 5.5 the default model with a 1M token context window. ·
New release v2.1.284 of anthropics/claude-code on GitHub.
191 points and 138 comments on Hacker News, linking anthropic.com.
MicroLLM Lab allows users to run seven small language models directly in web browsers using WebGPU. The tool supports chatting, benchmarking, and comparing models like PetitGPT and SmolLM2 on local devices. ·
Unsloth added decision model support, a document library, and Apple Silicon optimizations. The update also enables local Laya execution and accelerates image and video generation by up to 4.5 times. ·
A image to image model from Alissonerdx with 1,005 likes and 150.0k downloads on Hugging Face.
FuseReg improves representation autoencoders by training over random encoder layer subsets instead of fixed heuristics. This approach enhances both reconstruction quality and generation performance without requiring retraining. ·
Anthropic released Claude Code v2.1.284, making Claude Sonnet 5.5 the default model with a 1M token context window. ·
New release v2.1.284 of anthropics/claude-code on GitHub.
191 points and 138 comments on Hacker News, linking anthropic.com.
The author argues that the primary issue is a lack of understanding regarding system architecture and intent, rather than the use of AI to generate code. ·
Unsloth added decision model support, a document library, and Apple Silicon optimizations. The update also enables local Laya execution and accelerates image and video generation by up to 4.5 times. ·
FuseReg improves representation autoencoders by training over random encoder layer subsets instead of fixed heuristics. This approach enhances both reconstruction quality and generation performance without requiring retraining. ·
Researchers introduced RayOrch, a system for preparing foundation-model training data. It manages complex data pipelines by tracking relationships between parent items and their generated children. ·
| OpenAI | 1.9 |
| Nvidia | 1.43 |
| Opus 5.5 | 1.42 |
| LLM | 1.36 |
| Claude Sonnet 5.5 | 1.2 |
| Qwen3.8 | 0.98 |
Heat = summed hotness of today's stories naming the topic.
Releases, papers, trending weights and new listings
| convaiinnovations/laya | 4276 |
| Qwen/Qwen-Image-2.1 | 2576 |
| abenzerps/Qwen-Image-2.1-Uncensored-GGUF | 2231 |
| XingChen-AGI/Xing4.0-29B-A4B | 1798 |
| Edge0/Audio8-ASR-Infinite | 1368 |
| XingChen-AGI/TeleOCR | 720 |
| opus-5-5 | 91.15 |
| gpt-6-sol | 81.78 |
| deepseek-v4-flash-0731 | 62.82 |
Agents build our own projects from a brief; every run, failures included: https://cowerx.dev/daily/piratebench/
Vespper released a model fine-tuned for editing Word documents as an MCP. It reportedly makes agents three times faster and twice as cheap as the closest alternative. ·
19 points and 6 comments on Hacker News, linking vespper.com.
A text generation model from ukisai with 138 likes and 58.0k downloads on Hugging Face.
Unsloth added decision model support, a document library, and Apple Silicon optimizations. The update also enables local Laya execution and accelerates image and video generation by up to 4.5 times. ·
415 points and 375 comments on Hacker News, linking thecivilian.co.nz.
From Hugging Face blog.
Vespper released a model fine-tuned for editing Word documents as an MCP. It reportedly makes agents three times faster and twice as cheap as the closest alternative. ·
19 points and 6 comments on Hacker News, linking vespper.com.
Unsloth added decision model support, a document library, and Apple Silicon optimizations. The update also enables local Laya execution and accelerates image and video generation by up to 4.5 times. ·
415 points and 375 comments on Hacker News, linking thecivilian.co.nz.
189 points and 215 comments on Hacker News, linking platform.claude.com.
FuseReg improves representation autoencoders by training over random encoder layer subsets instead of fixed heuristics. This approach enhances both reconstruction quality and generation performance without requiring retraining. ·
Unsloth added decision model support, a document library, and Apple Silicon optimizations. The update also enables local Laya execution and accelerates image and video generation by up to 4.5 times. ·
New release v0.1.900-beta of unslothai/unsloth on GitHub.
415 points and 375 comments on Hacker News, linking thecivilian.co.nz.
Vespper released a model fine-tuned for editing Word documents as an MCP. It reportedly makes agents three times faster and twice as cheap as the closest alternative. ·
189 points and 215 comments on Hacker News, linking platform.claude.com.
FuseReg improves representation autoencoders by training over random encoder layer subsets instead of fixed heuristics. This approach enhances both reconstruction quality and generation performance without requiring retraining. ·
X posts other sources are linking to, then named voices on Bluesky and Mastodon
Law professor Orin Kerr on criminal liability for AI hacking
4 points and 0 comments on Hacker News, linking an X post.
I built a native Apple app for 4 platforms in one afternoon with my AI workflow
5 points and 2 comments on Hacker News, linking an X post.
Hospitals use AI to find more things to bill for. Insurers use AI to deny them.
10 points and 1 comments on Hacker News, linking an X post.
George Hotz’s opinion on AI coding
21 points and 10 comments on Hacker News, linking an X post.
Anthropic released Claude 5.5 Sonnet, which offers performance similar to Opus 5.5 at a lower cost. ·
Law professor Orin Kerr on criminal liability for AI hacking
4 points and 0 comments on Hacker News, linking an X post.
I built a native Apple app for 4 platforms in one afternoon with my AI workflow
5 points and 2 comments on Hacker News, linking an X post.
Hospitals use AI to find more things to bill for. Insurers use AI to deny them.
10 points and 1 comments on Hacker News, linking an X post.
George Hotz’s opinion on AI coding
21 points and 10 comments on Hacker News, linking an X post.
An agent in a loop optimizing a renderer with the goal to minimize frame times
2 points and 0 comments on Hacker News, linking an X post.
Law professor Orin Kerr on criminal liability for AI hacking
4 points and 0 comments on Hacker News, linking an X post.
I built a native Apple app for 4 platforms in one afternoon with my AI workflow
5 points and 2 comments on Hacker News, linking an X post.
Hospitals use AI to find more things to bill for. Insurers use AI to deny them.
10 points and 1 comments on Hacker News, linking an X post.
George Hotz’s opinion on AI coding
21 points and 10 comments on Hacker News, linking an X post.
New repos, releases and changelog
| yetone/magpie | 1572 |
| dzhng/jevgrep | 1314 |
| mikehasa/golive-skill | 1042 |
| ollaya-dev/ollaya | 851 |
| feitangyuan/onetake | 774 |
| yukitorido/short-video-generator-AI | 730 |
New release v1.9.0 of anthropics/anthropic-sdk-python on GitHub.
Unsloth added decision model support, a document library, and Apple Silicon optimizations. The update also enables local Laya execution and accelerates image and video generation by up to 4.5 times. ·
New release v3.20.0 of openai/openai-python on GitHub.
JoasASantos maintains a repository of AI models that are uncensored or fine-tuned for cybersecurity tasks. ·
OpenAI released version 0.158.0 of Codex, adding features for terminal interaction, OAuth connections, and image handling. The update also includes fixes for sandbox issues on Windows, Linux, and macOS. ·
Unsloth added decision model support, a document library, and Apple Silicon optimizations. The update also enables local Laya execution and accelerates image and video generation by up to 4.5 times. ·
New release v0.1.900-beta of unslothai/unsloth on GitHub.
OpenAI released version 0.158.0 of Codex, adding features for terminal interaction, OAuth connections, and image handling. The update also includes fixes for sandbox issues on Windows, Linux, and macOS. ·
JoasASantos maintains a repository of AI models that are uncensored or fine-tuned for cybersecurity tasks. ·
New release v1.9.0 of anthropics/anthropic-sdk-python on GitHub.
New release v3.20.0 of openai/openai-python on GitHub.
Unsloth added decision model support, a document library, and Apple Silicon optimizations. The update also enables local Laya execution and accelerates image and video generation by up to 4.5 times. ·
New release v0.1.900-beta of unslothai/unsloth on GitHub.
OpenAI released version 0.158.0 of Codex, adding features for terminal interaction, OAuth connections, and image handling. The update also includes fixes for sandbox issues on Windows, Linux, and macOS. ·
JoasASantos maintains a repository of AI models that are uncensored or fine-tuned for cybersecurity tasks. ·
A developer published a curated list of 180 open-source repositories. These tools allow coding agents to generate video and include security grades. ·
Prebuilt Linux x86_64 wheels for flash-attn, causal-conv1d, and mamba-ssm are available. They support CUDA 13, PyTorch 2.13 and 2.14, and Python 3.13. ·
Builds, power deals and the grid
| Tue 22 | 3 |
| Wed 23 | 3 |
| Thu 24 | 3 |
| Fri 25 | 20 |
| Sat 26 | 1 |
| Sun 27 | 1 |
| Mon 28 | 19 |
Stories from DCD, Data Center Knowledge, EIA and Hacker News, by publication day.
Goodman withdrew its application for a 90-megawatt data center campus in Sydney, citing significant changes to the local policy and regulatory environment. ·
The piece examines the design of silicon and systems required to support current and future data center workloads for artificial intelligence. ·
Supercapacitors offer millisecond response to AI power fluctuations. Experts view them as complements to existing battery and UPS systems rather than replacements. ·
10 points and 5 comments on Hacker News, linking wsj.com.
Goodman withdrew its application for a 90-megawatt data center campus in Sydney, citing significant changes to the local policy and regulatory environment. ·
The piece examines the design of silicon and systems required to support current and future data center workloads for artificial intelligence. ·
Supercapacitors offer millisecond response to AI power fluctuations. Experts view them as complements to existing battery and UPS systems rather than replacements. ·
10 points and 5 comments on Hacker News, linking wsj.com.
Goodman withdrew its application for a 90-megawatt data center campus in Sydney, citing significant changes to the local policy and regulatory environment. ·
The piece examines the design of silicon and systems required to support current and future data center workloads for artificial intelligence. ·
Supercapacitors offer millisecond response to AI power fluctuations. Experts view them as complements to existing battery and UPS systems rather than replacements. ·
GPUs, memory, and what tokens cost
| perceptron-mk1.5 | 1.5 |
| ember-1 | 15 |
| glm-5.3-prime | 8.8 |
| qwen3.8-max-prime | 12 |
| aion-3.5-mini | 1.4 |
| median, all paid models | 2 |
432 paid models listed; median of the 54 added in 30 days: $2.85. Rankings data by OpenRouter, CC BY 4.0.
Import AI issue 474 covers Platonic mindspace, space-based TPUs, and Zhipu's outer RSI loop. ·
From Import AI.
Import AI issue 474 covers Platonic mindspace, space-based TPUs, and Zhipu's outer RSI loop. ·
From Import AI.
Import AI issue 474 covers Platonic mindspace, space-based TPUs, and Zhipu's outer RSI loop. ·
From Import AI.
An experimental project provides a virtio device that allows KVM virtual machines to access NVIDIA GPUs with near-native performance. ·
