Apollo says AI has delivered 'zero financial payoff' outside the tech industry
8 points and 1 comments on Hacker News, linking businessinsider.com.
Top now
6 storiesWhat is climbing fastest across every source
LLMs7.4xOpenAI5.6xAnthropic5.7xrouterhubSI4.5xClaude2.4xClaude Code1.3x
What Security Teams Need to Know About Agents Using Abliterated Models
7 points and 0 comments on Hacker News, linking eve.security.
Nanbeige/Nanbeige4.3-33B
A text generation model from Nanbeige with 370 likes and 0 downloads on Hugging Face.
DenisSergeevitch/ex-ussr-kinship-research
A new repository written in Python with 536 stars in its first week.
Podcast: Kimi K3.1 Pro, the new cyber reality, and Muse musings
From Interconnects.
OpenAI and Anthropic cut their LLM prices by up to 80% within a few months. Why?
118 points and 138 comments on Hacker News, linking minimaxir.com.
Apollo says AI has delivered 'zero financial payoff' outside the tech industry
8 points and 1 comments on Hacker News, linking businessinsider.com.
Nanbeige/Nanbeige4.3-33B
A text generation model from Nanbeige with 370 likes and 0 downloads on Hugging Face.
DenisSergeevitch/ex-ussr-kinship-research
A new repository written in Python with 536 stars in its first week.
inclusionAI/Ming-3.0-tts-0.5B
A text to speech model from inclusionAI with 133 likes and 3,172 downloads on Hugging Face.
hevnsnt/CVEngine
A new repository written in Python with 135 stars in its first week.
A 10-year-old CNN outperforms frontier LLMs for skin cancer diagnosis
410 points and 199 comments on Hacker News, linking lynchlab.net.
AI use leads UC Berkeley math professor to see 'no point' in grading homework
195 points and 276 comments on Hacker News, linking dailycal.org.
DenisSergeevitch/ex-ussr-kinship-research
A new repository written in Python with 536 stars in its first week.
Nanbeige/Nanbeige4.3-33B
A text generation model from Nanbeige with 370 likes and 0 downloads on Hugging Face.
routerhubSI/uncensored-ai-api
A new repository with 365 stars in its first week.
anthropics/claude-code v2.1.310
Anthropic releases Claude Code version 2.1.310, which adds the Monitor tool to Bedrock, Vertex, and Foundry, and fixes several bugs. ·
Trending topics today: heat
Table
| LLMs | 2.17 |
| OpenAI | 1.39 |
| Anthropic | 0.83 |
| routerhubSI | 0.45 |
| Claude | 0.24 |
| Claude Code | 0.13 |
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 | 9336 |
| deepseek-ai/DeepSeek-V4.1-Flash-1017 | 2982 |
| black-forest-labs/flux-3-image-klein-comfy | 1604 |
| RecursiveSI/flint | 858 |
| XHToken/Spark-X2.6-4B ▲1 | 495 |
| Nanbeige/Nanbeige4.3-33B new | 370 |
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/
inclusionAI/Ming-3.0-tts-0.5B
A text to speech model from inclusionAI with 133 likes and 3,172 downloads on Hugging Face.
ttindall/Qwen3.8-27B-Defiance-Uncensored-MTP-GGUF
A text generation model from ttindall with 173 likes and 34.9k downloads on Hugging Face.
Cerebras Deploys Trillion-Parameter AI Models
350 points and 127 comments on Hacker News, linking cerebras.ai.
Five Qwen3.8-27B models match Fable 5 Using GVS5H Harness
182 points and 69 comments on Hacker News, linking github.com.
orcarouter/DeepSeek-V4.1-Flash-1017-abliterated-GGUF
A text generation model from orcarouter with 216 likes and 7,293 downloads on Hugging Face.
Cerebras Deploys Trillion-Parameter AI Models
350 points and 127 comments on Hacker News, linking cerebras.ai.
Five Qwen3.8-27B models match Fable 5 Using GVS5H Harness
182 points and 69 comments on Hacker News, linking github.com.
Securing our frontier model training environment
98 points and 99 comments on Hacker News, linking openai.com.
Dessa: Conversation Model trained only on speech
Dessa is a research preview of a full-duplex speech model that learns to talk by listening, mirroring how humans acquire language. ·
ISTA-DASLab/DeepSeek-V4.1-Flash-1017-GSQ-RCO-GGUF
A text generation model from ISTA-DASLab with 280 likes and 412.4k downloads on Hugging Face.
X voices
5 storiesX posts other sources are linking to, then named voices on Bluesky and Mastodon
@boneGPT
Imagine an AI future where being banned means you are now economically useless
94 points and 74 comments on Hacker News, linking an X post.
@mike64_t
Samsung running a TensorFlow lite model on 960x540 image to track you
285 points and 1 comments on Hacker News, linking an X post.
@chrislakin
Claude Code slyly adding users' emails to outbound HTTP
21 points and 5 comments on Hacker News, linking an X post.
@rynorhn
AI agents, in separate VMs without network, communicated through the CPU cache
2 points and 0 comments on Hacker News, linking an X post.
@mike64_t
Samsung running a TensorFlow lite model on 960x540 image to track you
285 points and 1 comments on Hacker News, linking an X post.
@boneGPT
Imagine an AI future where being banned means you are now economically useless
94 points and 74 comments on Hacker News, linking an X post.
@chrislakin
Claude Code slyly adding users' emails to outbound HTTP
21 points and 5 comments on Hacker News, linking an X post.
GitHub
5 storiesNew repos, releases and changelog
New AI repos this week: stars
Table
| wesbos/3d-model-anything | 1916 |
| rddmwmddr/android-ad-cleaner | 798 |
| dominikmartn/harp | 616 |
| udhaykumarbala/cleanyourtv-app | 581 |
| DenisSergeevitch/ex-ussr-kinship-research | 536 |
| Alejandros17487/ai-jailbreaks | 483 |
Jukebox-OS: Deterministic time-travel replay and ACID tool sandbox for AI agents
8 points and 0 comments on Hacker News, linking github.com.
Five Qwen3.8-27B models match Fable 5 Using GVS5H Harness
182 points and 69 comments on Hacker News, linking github.com.
hevnsnt/CVEngine
A new repository written in Python with 135 stars in its first week.
modelcontextprotocol/servers v2.1.0
New release v2.1.0 of modelcontextprotocol/servers on GitHub.
wangrongding/tv-fresh
A new repository written in Shell with 351 stars in its first week.
Five Qwen3.8-27B models match Fable 5 Using GVS5H Harness
182 points and 69 comments on Hacker News, linking github.com.
routerhubSI/uncensored-ai-api
A new repository with 365 stars in its first week.
wangrongding/tv-fresh
A new repository written in Shell with 351 stars in its first week.
rddmwmddr/android-ad-cleaner
A new repository written in Python with 798 stars in its first week.
routerhubSI/glm-5.3-abliterated
A new repository written in Python with 374 stars in its first week.
Infrastructure
5 storiesData centers, power, chips and what tokens cost
Data-center and energy stories per day
Table
| Tue 20 | 31 |
| Wed 21 | 29 |
| Thu 22 | 34 |
| Fri 23 | 39 |
| Sat 24 | 3 |
| Sun 25 | 1 |
| Mon 26 | 0 |
Stories from DCD, Data Center Knowledge, EIA and Hacker News, by publication day.
Median output price of models added in the last 30 days, $ per million tokens
One point per OpenRouter reading, last 8 days. Data by OpenRouter, CC BY 4.0.
Output price, $ per million tokens (newest models)
Table
| kimi-k3.1-pro | 10 |
| minimax-m3.1-flash | 0.6 |
| deepseek-v4.1-flash-1017 | 0.6 |
| glimpse | 3 |
| beam-1.0-501b-a23b | 1 |
| median, all paid models | 2 |
438 paid models listed; median of the 29 added in 30 days: $3.20. Rankings data by OpenRouter, CC BY 4.0.
Google illegally cut down 250K trees in Finland for data center
258 points and 209 comments on Hacker News, linking nypost.com.
GPU-unlock: FP/BF16 instruction issue rate on Nvidia 3000's cards
337 points and 80 comments on Hacker News, linking github.com.
AI:AM Highlights: Clean Neurons, Black October, the GPU Credit Crunch, and Buying the Supply Chain
Nathan Labenz and Prakash Narayanan host a podcast episode where guests Zvi Mowshowitz, Evan Conrad, and Lonis Hamaili discuss AI development constraints, including credit availability and supply chain value. ·
Google illegally cut down 250K trees in Finland for data center
258 points and 209 comments on Hacker News, linking nypost.com.
GPU-unlock: FP/BF16 instruction issue rate on Nvidia 3000's cards
337 points and 80 comments on Hacker News, linking github.com.
Google will pay people to use less electricity so its data centers can use more
Google pays people to reduce electricity use so its data centers can access more power. The company is building a 1GW virtual power plant. ·
TornadoVM 7 Enables GPU-Native Workloads from Java Applications
85 points and 17 comments on Hacker News, linking infoq.com.