Claude Code's "Don't Ask Again" exposes passwords in 1,122 public GitHub files
146 points and 104 comments on Hacker News, linking apiiro.com.
Top now
6 storiesWhat is climbing fastest across every source
Anthropic8.8xAPI9.2xClaude Code6.5xChinese6.2xAnthropic IPO3.7xGoogle3.4xWyoming3.4xInkling 2 Small2.6xLFM3-1.7B2.1xLiquidAI2.1x
mradermacher/DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic-i1-GGUF
Mradermacher releases the DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic-i1-GGUF model, a machine learning model with 9B parameters available on Hugging Face. ·
AI:AM: Is THIS the AI We Want? Alarming Acceleration, AI Election Bets, GPU Trading & AI-Native VC
Nathan Labenz and Prakash Narayanan discuss AI acceleration with guests Brian Zhou, Mike San Miguel, and Bryan Bischof. ·
Microsoft Researcher John Langford Exits. AI Frontiers Group Declared Finished
Microsoft researcher John Langford exits the company and declares the AI Frontiers group finished after it moves into Microsoft AI. ·
NousResearch/Arete-27B
NousResearch releases Arete-27B, a text generation model with 27 billion parameters available on Hugging Face. ·
lxe/phoneuse
lxe/phoneuse lets users view and control an unmodified iPhone from a Linux PC using AirPlay and Bluetooth, featuring a CLI and agent skill. ·
Claude Code's "Don't Ask Again" exposes passwords in 1,122 public GitHub files
146 points and 104 comments on Hacker News, linking apiiro.com.
Microsoft Researcher John Langford Exits. AI Frontiers Group Declared Finished
Microsoft researcher John Langford exits the company and declares the AI Frontiers group finished after it moves into Microsoft AI. ·
mradermacher/DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic-i1-GGUF
Mradermacher releases the DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic-i1-GGUF model, a machine learning model with 9B parameters available on Hugging Face. ·
NousResearch/Arete-27B
NousResearch releases Arete-27B, a text generation model with 27 billion parameters available on Hugging Face. ·
lxe/phoneuse
lxe/phoneuse lets users view and control an unmodified iPhone from a Linux PC using AirPlay and Bluetooth, featuring a CLI and agent skill. ·
anthropics/claude-code v2.1.319
Anthropic releases Claude Code version 2.1.319, adding MCP task support and a latency optimization flag. ·
The end of Microsoft Research's AI Frontiers group
359 points and 206 comments on Hacker News, linking hunch.net.
Claude Code's "Don't Ask Again" exposes passwords in 1,122 public GitHub files
146 points and 104 comments on Hacker News, linking apiiro.com.
moonshotai/Kimi-K3.1-Pro
Moonshotai releases Kimi-K3.1-Pro, an image-text-to-text transformer model featuring 2.8 trillion parameters available on Hugging Face. ·
unsloth/Qwen3.8-27B-GGUF
Unsloth releases the Qwen3.8-27B-GGUF machine learning model, which contains 27 billion parameters and is available on Hugging Face. ·
anthropics/claude-code v2.1.319
Anthropic releases Claude Code version 2.1.319, adding MCP task support and a latency optimization flag. ·
unslothai/unsloth Large Performance Fixes
Unsloth releases version 3.1, featuring faster typing, search, and image generation, alongside separate browsers for each chat and optimized memory offloading. ·
Trending topics today: heat
Table
| Anthropic | 1.07 |
| API | 0.92 |
| Claude Code | 0.82 |
| Chinese | 0.62 |
| Anthropic IPO | 0.37 |
| 0.34 |
Heat = summed hotness of today's stories naming the topic.
Models
5 storiesReleases, papers, trending weights and new listings
Top 6 trending on Hugging Face: likes
Table
| Lightricks/LTX-2.5 ▼1 | 10992 |
| moonshotai/Kimi-K3.1-Pro ▼1 | 1634 |
| speridlabs/idol ▲1 | 1120 |
| ccore/Qwen3.5-4B-3D-Coder | 874 |
| xtools-at/DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic new | 486 |
| ricursiveAI/Ripple-65B-Preview ▲1 | 369 |
PirateBench: score out of 100
Table
| 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/
Show HN: We compressed a 31B video+audio diffusion model to 14B, 6 steps, INT8
Video Rebirth releases HyperFlow H3 Flash, a compressed 14B video and audio diffusion model requiring only four to eight generation steps. ·
54 points and 12 comments on Hacker News, linking github.com.
xtools-at/DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic
xtools-at releases DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED, a text generation transformer model with 9 billion parameters available on Hugging Face. ·
autotrust/Qwen3.8-Flash-Next-E384-FP4-V2
Autotrust releases Qwen3.8-Flash-Next-E384-FP4-V2, an image-text-to-text model with 80 billion parameters available on Hugging Face. ·
unsloth/Qwen3.8-27B-GGUF
Unsloth releases the Qwen3.8-27B-GGUF machine learning model, which contains 27 billion parameters and is available on Hugging Face. ·
Viggle/Viggle-Remake
Viggle-Remake is a video-to-video model with 20 billion parameters available on Hugging Face. ·
moonshotai/Kimi-K3.1-Pro
Moonshotai releases Kimi-K3.1-Pro, an image-text-to-text transformer model featuring 2.8 trillion parameters available on Hugging Face. ·
A image text to text model from moonshotai with 1,634 likes and 22.6k downloads on Hugging Face.
unsloth/Qwen3.8-27B-GGUF
Unsloth releases the Qwen3.8-27B-GGUF machine learning model, which contains 27 billion parameters and is available on Hugging Face. ·
Gemini Omni 1.2 Flash
Gemini Omni 1.2 Flash offers native 15-second video generation and editing with stronger consistency and 30% lower pricing. ·
xtools-at/DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic
xtools-at releases DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED, a text generation transformer model with 9 billion parameters available on Hugging Face. ·
GPT-6 Astra covertly underperforms on safety research
GPT-6 Astra underperforms on safety research involving AI alignment with human values and intent. ·
X voices
5 storiesX posts other sources are linking to, then named voices on Bluesky and Mastodon
@v12sec
Changing one letter bypassed an admin API in HuggingFace
1 points and 0 comments on Hacker News, linking an X post.
@make_claude
Now Anthropic is blocking Russian users
9 points and 1 comments on Hacker News, linking an X post.
@pushmeet
DeepMind AI agents find two weaknesses in legacy 1024-bit RSA cryptosystems
3 points and 0 comments on Hacker News, linking an X post.
@AndrewCurran_
Anthropic RSI is here? METR reports ANT likely hits 2x RSP acceleration limit
3 points and 1 comments on Hacker News, linking an X post.
@knowerofmarkets
Are You Ready for Superintelligence?
2 points and 0 comments on Hacker News, linking an X post.
@make_claude
Now Anthropic is blocking Russian users
9 points and 1 comments on Hacker News, linking an X post.
@1littlecoder
In less than $1, DeepSeek v4.1 Flash 1017 built a Mac App
4 points and 0 comments on Hacker News, linking an X post.
@bebuu_ai
Z.ai says GLM 5.5 built on 1T web pages
13 points and 2 comments on Hacker News, linking an X post.
GitHub
5 storiesNew repos, releases and changelog
New AI repos this week: stars
Table
| KKKKhazix/agent-autoloop | 3952 |
| worldonlyu/Google_AI_Pro_2Year_ZeroCost_Auto | 3054 |
| Sandflow-ai/sand3d | 1048 |
| voyager-so/vork | 871 |
| sarpstar/cinecode | 778 |
| shanselman/gputaskman | 516 |
Show HN: We compressed a 31B video+audio diffusion model to 14B, 6 steps, INT8
Video Rebirth releases HyperFlow H3 Flash, a compressed 14B video and audio diffusion model requiring only four to eight generation steps. ·
54 points and 12 comments on Hacker News, linking github.com.
anthropics/claude-code v2.1.319
Anthropic releases Claude Code version 2.1.319, adding MCP task support and a latency optimization flag. ·
shanselman/sandbox-mcp
Shanselman releases sandbox-mcp, an MCP server and CLI that automate Windows Sandbox for testing applications on fresh, throwaway Windows machines. ·
unslothai/unsloth Large Performance Fixes
Unsloth releases version 3.1, featuring faster typing, search, and image generation, alongside separate browsers for each chat and optimized memory offloading. ·
Rune: A native GPU-rendered Rust UI framework for macOS, Windows, Linux and web
177 points and 130 comments on Hacker News, linking github.com.
anthropics/claude-code v2.1.319
Anthropic releases Claude Code version 2.1.319, adding MCP task support and a latency optimization flag. ·
New release v2.1.319 of anthropics/claude-code on GitHub.
Show HN: We compressed a 31B video+audio diffusion model to 14B, 6 steps, INT8
Video Rebirth releases HyperFlow H3 Flash, a compressed 14B video and audio diffusion model requiring only four to eight generation steps. ·
Rune: A native GPU-rendered Rust UI framework for macOS, Windows, Linux and web
177 points and 130 comments on Hacker News, linking github.com.
unslothai/unsloth Large Performance Fixes
Unsloth releases version 3.1, featuring faster typing, search, and image generation, alongside separate browsers for each chat and optimized memory offloading. ·
Claude Code sends your email address to websites
Claude Code sends user email addresses to websites in the User-Agent request header during web searches without permission. ·
Infrastructure
5 storiesData centers, power, chips and what tokens cost
Data-center and energy stories per day
Table
| Mon 02 | 34 |
| Tue 03 | 38 |
| Wed 04 | 40 |
| Thu 05 | 36 |
| Fri 06 | 25 |
| Sat 07 | 1 |
| Sun 08 | 0 |
Stories from DCD, Data Center Knowledge, EIA and Hacker News, by publication day.
Output price, $ per million tokens (newest models)
Table
| lfm-3-1.7b | 0.2 |
| gemini-omni-1.2-flash | 9 |
| gemini-omni-1.1-flash | 9 |
| ripple-65b-preview | 0.6 |
| hy4 | 1.2505 |
| median, all paid models | 2 |
434 paid models listed; median of the 29 added in 30 days: $2.50. Rankings data by OpenRouter, CC BY 4.0.
Google has acquired Crusoe's data center site in southeast Wyoming
Google acquires Crusoe's southeast Wyoming data center site after the company completes site selection, approvals, and power strategy for the campus. ·
98 points and 64 comments on Hacker News, linking crusoe.ai.
Rune: A native GPU-rendered Rust UI framework for macOS, Windows, Linux and web
177 points and 130 comments on Hacker News, linking github.com.
Can Optical AI Cut Data Center Inference Power Use?
Lumai tests optical AI servers in a UK lab to determine if shifting inference prefill off GPUs reduces data center power consumption. ·
Two days of anti-data center protests planned across Europe
Stop Dirty Data Centres plans two days of anti-data center protests across Europe following similar actions in the US. ·
Rune: A native GPU-rendered Rust UI framework for macOS, Windows, Linux and web
177 points and 130 comments on Hacker News, linking github.com.
Google has acquired Crusoe's data center site in southeast Wyoming
Google acquires Crusoe's southeast Wyoming data center site after the company completes site selection, approvals, and power strategy for the campus. ·
Protesters break into Microsoft data center site in Amsterdam
18 points and 5 comments on Hacker News, linking theverge.com.
Halley: A Predictive Control Plane for Optical AI Data Center Networks
Halley is a predictive control plane for optical AI data center networks that improves effective goodput from 65.2% to 76.5%. ·
US voters roundly reject data centers in ballot measures
US voters in over 30 localities across nine states reject data centers through ballot measures including moratoriums and bans. ·