The avalanche that hit math is coming for AI development
An avalanche hits math, which now impacts AI development, leading to an expectation of superintelligence by 2030. ·
44 points and 45 comments on Hacker News, linking dwarkesh.com.
The front page of the internet for AI users · Fri 13 Nov 2026, 20:00 UTC · Get it by email
What is climbing fastest across every source
An avalanche hits math, which now impacts AI development, leading to an expectation of superintelligence by 2030. ·
44 points and 45 comments on Hacker News, linking dwarkesh.com.
HauhauCS/Qwen3.8-27B-Uncensored-HauhauCS-Aggressive-MTP-GGUF is an image-text-to-text model with 2B parameters available on Hugging Face. ·
5 points and 0 comments on Hacker News, linking perplexity.ai.
The qwen-concept-27b model is a machine learning model with 27 billion parameters available on Hugging Face. ·
GitHub Copilot app now supports GitHub Enterprise Server, allowing users to add their GHES account for remote repositories, issues, and pull requests. ·
From Gary Marcus.
An avalanche hits math, which now impacts AI development, leading to an expectation of superintelligence by 2030. ·
44 points and 45 comments on Hacker News, linking dwarkesh.com.
Models can transmit viruses through weight space, where direct modification of neural network weights changes knowledge, behavior, and alignment. ·
HauhauCS/Qwen3.8-27B-Uncensored-HauhauCS-Aggressive-MTP-GGUF is an image-text-to-text model with 2B parameters available on Hugging Face. ·
The qwen-concept-27b model is a machine learning model with 27 billion parameters available on Hugging Face. ·
Ronni35/ai--manus describes Manus, a general-purpose AI agent system that runs tools and operations in a sandbox environment. ·
A new repository with 155 stars in its first week.
An avalanche hits math, which now impacts AI development, leading to an expectation of superintelligence by 2030. ·
44 points and 45 comments on Hacker News, linking dwarkesh.com.
The Gemini Decisions API uses the Gemini Decision model to provide fast, calibrated multimodal decisions. ·
HauhauCS/Qwen3.8-27B-Uncensored-HauhauCS-Aggressive-MTP-GGUF is an image-text-to-text model with 2B parameters available on Hugging Face. ·
The qwen-concept-27b model is a machine learning model with 27 billion parameters available on Hugging Face. ·
ggml-org releases llama.cpp version 1.0.0, which includes a nightly build labeled b12359 and a changelog since version 0.9.0. ·
SO-Arena benchmarks System One models across 11 domains and five capability families, assessing decision reliability through stability and missing-answer recognition. ·
| Jensen Huang | 2.59 |
| NVIDIA | 2.14 |
| Ronni35 | 1.56 |
| xombiensar | 1.56 |
| Floxaep1 | 1.49 |
| Gemini 4 Hydrogen | 1.47 |
Heat = summed hotness of today's stories naming the topic.
Releases, papers, trending weights and new listings
| Qwen/Qwen3.8-27B ▲2 | 19845 |
| Lightricks/LTX-2.5 ▲1 | 11749 |
| xtools-at/DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic | 5207 |
| speridlabs/idol ▼2 | 2279 |
| MiniMaxAI/MiniMax-H3-Image | 987 |
| NousResearch/Arete-27B | 530 |
| gpt-6.1-sol | 98.9 |
| opus-5-5 | 96.46 |
| gpt-6-sol | 91.6 |
| gpt-6-luna | 65.25 |
| haiku-4-5 | 4.28 |
Agents build our own projects from a brief; every run, failures included: https://cowerx.dev/daily/piratebench/
Models can transmit viruses through weight space, where direct modification of neural network weights changes knowledge, behavior, and alignment. ·
60 points and 28 comments on Hacker News, linking alphaxiv.org.
An author conducts 600 experiments to analyze JEPA world models, contrasting them with generative models that predict the future by generating pixels. ·
OpenAI disbands its robotics team and moves researchers to other groups as its general models begin running robots. ·
Claude's status page reports elevated errors on claude.ai, Claude Code, and platform.claude.com. ·
191 points and 164 comments on Hacker News, linking status.claude.com.
The Gemini Decisions API uses the Gemini Decision model to provide fast, calibrated multimodal decisions. ·
565 points and 161 comments on Hacker News, linking blog.google.
Claude's status page reports elevated errors on claude.ai, Claude Code, and platform.claude.com. ·
SO-Arena benchmarks System One models across 11 domains and five capability families, assessing decision reliability through stability and missing-answer recognition. ·
191 points and 164 comments on Hacker News, linking status.claude.com.
An author conducts 600 experiments to analyze JEPA world models, contrasting them with generative models that predict the future by generating pixels. ·
X posts other sources are linking to, then named voices on Bluesky and Mastodon
37% on ARC-AGI-2 Kaggle Without LLMs
5 points and 0 comments on Hacker News, linking an X post.
For my first post, I'm sharing a letter Nvidia signed on why open models matter
3 points and 2 comments on Hacker News, linking an X post.
Forensic evidence Chinese AI labs built their newest models on Nvidia Blackwell
6 points and 0 comments on Hacker News, linking an X post.
Nvidia prepares to fire CEO Jensen Huang, WSJ reports
225 points and 4 comments on Hacker News, linking an X post.
Reprogramming Dormant Neurons to Backdoor AI Agents
6 points and 1 comments on Hacker News, linking an X post.
Nvidia prepares to fire CEO Jensen Huang, WSJ reports
225 points and 4 comments on Hacker News, linking an X post.
Forensic evidence Chinese AI labs built their newest models on Nvidia Blackwell
6 points and 0 comments on Hacker News, linking an X post.
For my first post, I'm sharing a letter Nvidia signed on why open models matter
3 points and 2 comments on Hacker News, linking an X post.
Reprogramming Dormant Neurons to Backdoor AI Agents
6 points and 1 comments on Hacker News, linking an X post.
Fake accounts defended Jensen Huang and attacked lawmakers over chip exports
10 points and 0 comments on Hacker News, linking an X post.
New repos, releases and changelog
| HybridRobotics/supermimic | 1422 |
| agentes-chat/agentes-chat | 1323 |
| FrigadeHQ/yogurt | 1222 |
| dabingfy-dot/solo-to-team-skills | 841 |
| alexeyw/lot-map | 713 |
| lxe/phoneuse | 648 |
Ronni35/ai--manus describes Manus, a general-purpose AI agent system that runs tools and operations in a sandbox environment. ·
A new repository with 157 stars in its first week.
A new repository with 155 stars in its first week.
Obasideyvid/ai--manus describes Manus, a general-purpose AI agent system that runs tools and operations in a sandbox environment. ·
A new repository with 154 stars in its first week.
xombiensar/manus--ai describes AI Manus, a general-purpose AI agent system that runs tools and operations in a sandbox environment. ·
ggml-org releases llama.cpp version 1.0.0, which includes a nightly build labeled b12359 and a changelog since version 0.9.0. ·
New release v1.0.0 of ggml-org/llama.cpp on GitHub.
OpenAI Codex version 0.169.0 adds multiple side-chat tabs, custom model providers, and Cloud connection support to the terminal workspace. ·
Anthropic releases Claude Code version 2.1.323, adding Code Mode, plugin validation, and .docx support. ·
OpenAI Codex version 0.168.0 adds model provider selection, side conversations, audio recording support, animated GIF generation, and terminal notifications. ·
HyperionaiDev creates smolmaxxing to maximize the ability of running different large language models on small and old devices. ·
Data centers, power, chips and what tokens cost
| Sat 07 | 1 |
| Sun 08 | 2 |
| Mon 09 | 35 |
| Tue 10 | 31 |
| Wed 11 | 31 |
| Thu 12 | 29 |
| Fri 13 | 29 |
Stories from DCD, Data Center Knowledge, EIA and Hacker News, by publication day.
| gemini-4-hydrogen | 7.5 |
| gemini-4-hydrogen:batch | 3.75 |
| ernie-light-dt-1.0-intl | 3.94 |
| gemini-4-carbon | 10 |
| gemini-4-carbon:batch | 5 |
| median, all paid models | 2.04 |
444 paid models listed; median of the 36 added in 30 days: $2.50. Rankings data by OpenRouter, CC BY 4.0.
Forensic analysis of public AI models reveals that DeepSeek, Moonshot AI, MiniMax, and Ant Group use NVIDIA Blackwell GPUs to train their systems. ·
349 points and 144 comments on Hacker News, linking unemploymentlabs.com.
Data centers pay 38% of energy infrastructure costs in 2025, while rural and low-income households cover the rest. ·
Forensic analysis of public AI models reveals that DeepSeek, Moonshot AI, MiniMax, and Ant Group use NVIDIA Blackwell GPUs to train their systems. ·
349 points and 144 comments on Hacker News, linking unemploymentlabs.com.
Data centers pay 38% of energy infrastructure costs in 2025, while rural and low-income households cover the rest. ·
583 points and 275 comments on Hacker News, linking michael.bacarella.com.