The front page of the internet for AI users · Sun 01 Nov 2026, 01:00 UTC · Get it by email
What is climbing fastest across every source
A team releases Memdoor, a command-line coding and automation agent that runs locally or over SSH and uses five Qwen3.8-27B models to beat Fable 5 on LiveCodeBench Hard. ·
50 points and 23 comments on Hacker News, linking github.com.
Debate has arisen over whether watermarking AI text output, now being adopted by major labs under EU AI Act transparency guidelines, affects the quality of the text models produce. ·
A text-generation model called Qwen3.5-4B-Qwenputer is available on Hugging Face for download and running. The model has 9B parameters. ·
Parameter Labs describes an exploit chain against ChatGPT's HTML preview sandbox, where a MathLive keyboard message handler using innerHTML let untrusted iframe code inject an unsandboxed iframe. ·
Datasette releases version 0.1a9 of its datasette-llm plugin, which now uses the default model set by LLM so Datasette Agent can run without configuring a model. ·
From Cognitive Revolution.
A team releases Memdoor, a command-line coding and automation agent that runs locally or over SSH and uses five Qwen3.8-27B models to beat Fable 5 on LiveCodeBench Hard. ·
50 points and 23 comments on Hacker News, linking github.com.
Anthropic takes steps to block its Claude model from reverse-engineering video games, and the restrictions affect legitimate game developers and basic troubleshooting tasks. ·
A new repository written in TypeScript with 2,389 stars in its first week.
Tavern Labs AI offers a character AI SDK that bundles default personas, persistent character memory, and provider adapters in one package. ·
Anthropic releases Claude Code v2.1.316, which adds settings to disable the ultrathink keyword and to set subagent effort levels, plus pooled group budgets for the Claude apps gateway. ·
A text-to-speech model called Audio8-TTS-Realistic-0.6b is available on Hugging Face for download and use. The model has 0.6 billion parameters. ·
282 points and 144 comments on Hacker News, linking github.com.
A new repository written in TypeScript with 2,389 stars in its first week.
GPT-5.6 Sol, running in Codex at a low thinking level, completes a peaceful Factorio run in 13.5 hours of game time using a Python game library. ·
Anthropic releases Claude Code v2.1.316, which adds settings to disable the ultrathink keyword and to set subagent effort levels, plus pooled group budgets for the Claude apps gateway. ·
Capri is a fan-made cyberpunk AI agent for Windows that controls files and code via Claude Code, Codex, or Antigravity without requiring a terminal. ·
Unsloth updates its platform with custom backgrounds, faster app updates, improved Linux performance, and DeepSeek-V4.1-Flash-1017 support. ·
| Anthropic | 3.84 |
| Claude | 2.02 |
| LLMs | 0.91 |
| IPO | 0.58 |
| tavern-labs-ai | 0.44 |
| Claude Code | 0.29 |
Heat = summed hotness of today's stories naming the topic.
Releases, papers, trending weights and new listings
| deepseek-ai/DeepSeek-V4.1-Flash-1017 ▼1 | 4103 |
| mistralai/Mistral-Large-4-1T-A52B ▲2 | 1403 |
| Nanbeige/Nanbeige4.3-33B ▲1 | 874 |
| LiquidAI/LFM3-1.7B | 633 |
| nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B | 545 |
| baidu/Unlimited-OCR-2 | 473 |
| 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/
Anthropic takes steps to block its Claude model from reverse-engineering video games, and the restrictions affect legitimate game developers and basic troubleshooting tasks. ·
48 points and 33 comments on Hacker News, linking mashable.com.
GPT-5.6 Sol, running in Codex at a low thinking level, completes a peaceful Factorio run in 13.5 hours of game time using a Python game library. ·
44 points and 30 comments on Hacker News, linking datavorous.com.
52 points and 1 comments on Hacker News, linking github.com.
A new repository written in TypeScript with 2,389 stars in its first week.
282 points and 144 comments on Hacker News, linking github.com.
A new repository written in TypeScript with 2,389 stars in its first week.
GPT-5.6 Sol, running in Codex at a low thinking level, completes a peaceful Factorio run in 13.5 hours of game time using a Python game library. ·
Mistral releases Mistral Large 4, a one trillion parameter multimodal model trained in Europe, as the best open-weight coding model. ·
Anthropic takes steps to block its Claude model from reverse-engineering video games, and the restrictions affect legitimate game developers and basic troubleshooting tasks. ·
X posts other sources are linking to, then named voices on Bluesky and Mastodon
Anthropic blocks users for anything touching training larger models
173 points and 78 comments on Hacker News, linking an X post.
Anthropic: Scaling Misuse Detection with Self-Assessment of Chats
7 points and 0 comments on Hacker News, linking an X post.
How this AI Agent ended +$517.61 in profits during a 10% market sell-off
5 points and 0 comments on Hacker News, linking an X post.
Anthropic Cracking Down on Video Game Rewrites
124 points and 127 comments on Hacker News, linking an X post.
Reddit CEO admits Google's AI Mode is choking off its search traffic
8 points and 7 comments on Hacker News, linking an X post.
Anthropic blocks users for anything touching training larger models
173 points and 78 comments on Hacker News, linking an X post.
Anthropic Cracking Down on Video Game Rewrites
124 points and 127 comments on Hacker News, linking an X post.
Building an AI-native engineering org
184 points and 194 comments on Hacker News, linking an X post.
Anthropic: Scaling Misuse Detection with Self-Assessment of Chats
7 points and 0 comments on Hacker News, linking an X post.
How this AI Agent ended +$517.61 in profits during a 10% market sell-off
5 points and 0 comments on Hacker News, linking an X post.
New repos, releases and changelog
| anhao/ai-pose-gallery-zh | 4677 |
| heydamir/unslop | 3487 |
| wangzan101/Mika | 2733 |
| deepseek-ai/dsh-superagent-prototype | 2389 |
| paipiange/nvzhuang | 1718 |
| andyhuo520/digital-human-koubo | 1587 |
52 points and 1 comments on Hacker News, linking github.com.
A new repository written in TypeScript with 2,389 stars in its first week.
282 points and 144 comments on Hacker News, linking github.com.
Tavern Labs AI offers a character AI SDK that bundles default personas, persistent character memory, and provider adapters in one package. ·
Anthropic releases Claude Code v2.1.316, which adds settings to disable the ultrathink keyword and to set subagent effort levels, plus pooled group budgets for the Claude apps gateway. ·
282 points and 144 comments on Hacker News, linking github.com.
A new repository written in TypeScript with 2,389 stars in its first week.
Anthropic releases Claude Code v2.1.316, which adds settings to disable the ultrathink keyword and to set subagent effort levels, plus pooled group budgets for the Claude apps gateway. ·
Capri is a fan-made cyberpunk AI agent for Windows that controls files and code via Claude Code, Codex, or Antigravity without requiring a terminal. ·
Tavern Labs releases the Persona AI SDK, a package that provides default personas, durable character memory, and adapters for Bedrock and OpenAI. ·
Data centers, power, chips and what tokens cost
| Mon 26 | 39 |
| Tue 27 | 34 |
| Wed 28 | 32 |
| Thu 29 | 42 |
| Fri 30 | 29 |
| Sat 31 | 2 |
| Sun 01 | 0 |
Stories from DCD, Data Center Knowledge, EIA and Hacker News, by publication day.
One point per OpenRouter reading, last 8 days. Data by OpenRouter, CC BY 4.0.
| particle-1 | 0.5 |
| gpt-6.1-terra-pro | 5 |
| gpt-6.1-terra-pro:batch | 2.5 |
| gpt-6.1-terra | 5 |
| gpt-6.1-terra:batch | 2.5 |
| median, all paid models | 2 |
426 paid models listed; median of the 29 added in 30 days: $2.50. Rankings data by OpenRouter, CC BY 4.0.
Europe's growing data center demand is outpacing its outdated electricity grid, limiting the region's ability to use renewable energy for AI expansion. ·
From DCD.
A Scottish automotive company proposes a data center near Cowlairs Park in Glasgow and is consulting the local council about the project. ·
From DCD.
34 points and 31 comments on Hacker News, linking youtube.com.
SkyPilot optimizes NVLink usage in NVL72 racks through topology-aware placement, in-place recovery, and automatic defragmentation. ·
Europe's growing data center demand is outpacing its outdated electricity grid, limiting the region's ability to use renewable energy for AI expansion. ·
A Scottish automotive company proposes a data center near Cowlairs Park in Glasgow and is consulting the local council about the project. ·