OpenAI just dropped 700 preprints of mathematical proofs and counterexamples
22 points and 6 comments on Hacker News, linking github.com.
Top now
6 storiesWhat is climbing fastest across every source
Mistral Large 47.2xOpenAI7.5xLe Chonk7.1xChatGPT6.7xEmbeddingGemma 27.2xgoogle7.0xClaude5.9xMeta8.5xReflection7.1xMistral7.0x
Penguin Mail – open-source Rust email client for Linux with AI
Penguin Mail is an open-source Rust email client for Linux that integrates AI with Gmail, Microsoft, and other accounts, plus calendar and contacts, in one local application. ·
Why Jev Is Changing How We Build With AI with Diogo Almeida - #779
TypeSafe co-founder Diogo Almeida explains that Jev, the company's new model, uses reinforcement learning to provide reliable intelligence for software automation. ·
llm-mistral 0.16
The llm-mistral plugin version 0.16 adds support for reasoning models, including the newly released Mistral Large 4. ·
openai/openai-python v3.26.0
OpenAI releases version 3.26.0 of its Python library, which adds standalone Decisions support to the API. ·
anthropics/claude-code v2.1.292
New release v2.1.292 of anthropics/claude-code on GitHub.
Mistral Large 4: "Le Chonk"
Mistral Large 4, nicknamed Le Chonk, is an enterprise AI platform that allows users to customize, fine-tune, and deploy assistants, agents, and multimodal models. ·
518 points and 5 comments on Hacker News, linking mistral.ai.
EmbeddingGemma 2: An open, lightweight multimodal embedding model
191 points and 27 comments on Hacker News, linking blog.google.
anthropics/claude-code v2.1.292
New release v2.1.292 of anthropics/claude-code on GitHub.
google/embeddinggemma-2
A feature extraction model from google with 426 likes and 364 downloads on Hugging Face.
unslothai/unsloth New Browser + Voice Cloning
Unsloth releases a new version featuring a built-in browser panel and support for Google's EmbeddingGemma 2 model. ·
Kandinsky 6.0 Video: Foundation Models for Synchronized Video and Audio Generation
Kandinsky 6.0 Video is a family of diffusion models that generates synchronized text-to-audio-video clips, with the Pro version containing 29 billion parameters. ·
Mistral Large 4: "Le Chonk"
Mistral Large 4, nicknamed Le Chonk, is an enterprise AI platform that allows users to customize, fine-tune, and deploy assistants, agents, and multimodal models. ·
518 points and 5 comments on Hacker News, linking mistral.ai.
Sharing AI Progress in Mathematics
OpenAI publishes new results on open mathematical problems from an internal frontier model and shares Lean proof formalizations on GitHub. ·
anthropics/claude-code v2.1.292
New release v2.1.292 of anthropics/claude-code on GitHub.
google/embeddinggemma-2
A feature extraction model from google with 426 likes and 364 downloads on Hugging Face.
unslothai/unsloth New Browser + Voice Cloning
Unsloth releases a new version featuring a built-in browser panel and support for Google's EmbeddingGemma 2 model. ·
Kandinsky 6.0 Video: Foundation Models for Synchronized Video and Audio Generation
Kandinsky 6.0 Video is a family of diffusion models that generates synchronized text-to-audio-video clips, with the Pro version containing 29 billion parameters. ·
Trending topics today: heat
Table
| Mistral Large 4 | 3.47 |
| OpenAI | 1.91 |
| Le Chonk | 1.9 |
| ChatGPT | 1.57 |
| EmbeddingGemma 2 | 1.31 |
| 1.26 |
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
| convaiinnovations/laya ▼4 | 5279 |
| abenzerps/Qwen-Image-2.1-Uncensored-GGUF | 3424 |
| Cloudflare/clef | 1672 |
| autotrust/JEV-27B-VL ▲5 | 974 |
| Aleph-Alpha/Kolibri-1 | 711 |
| Cloudflare/clef-flash | 590 |
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/
Mistral Large 4: "Le Chonk"
Mistral Large 4, nicknamed Le Chonk, is an enterprise AI platform that allows users to customize, fine-tune, and deploy assistants, agents, and multimodal models. ·
518 points and 5 comments on Hacker News, linking mistral.ai.
EmbeddingGemma 2: An open, lightweight multimodal embedding model
191 points and 27 comments on Hacker News, linking blog.google.
google/embeddinggemma-2
A feature extraction model from google with 426 likes and 364 downloads on Hugging Face.
Claude Code’s suggested message feature: I think the real customer is the model
69 points and 38 comments on Hacker News, linking zohaib.cc.
Mistral Large 4
Mistral Large 4 is an open-weight multimodal model using a Mixture-of-Experts design. It contains 49 billion active parameters and 1.05 trillion total parameters. ·
Mistral Large 4: "Le Chonk"
Mistral Large 4, nicknamed Le Chonk, is an enterprise AI platform that allows users to customize, fine-tune, and deploy assistants, agents, and multimodal models. ·
518 points and 5 comments on Hacker News, linking mistral.ai.
EmbeddingGemma 2: An open, lightweight multimodal embedding model
191 points and 27 comments on Hacker News, linking blog.google.
google/embeddinggemma-2
A feature extraction model from google with 426 likes and 364 downloads on Hugging Face.
Mistral Large 4
Mistral Large 4 is an open-weight multimodal model using a Mixture-of-Experts design. It contains 49 billion active parameters and 1.05 trillion total parameters. ·
Mistral Large 4
1,546 points and 945 comments on Hacker News, linking mistral.ai.
X voices
5 storiesX posts other sources are linking to, then named voices on Bluesky and Mastodon
@Frances01896069
UT Austin Math Chair says OpenAI preparing to release 400 AI-generated proofs
2 points and 0 comments on Hacker News, linking an X post.
@mxfp0
Cloudflare's decision model "Clef" playing Balatro
1 points and 0 comments on Hacker News, linking an X post.
@backnotprop
Decision making is becoming the agentic coding bottleneck
2 points and 0 comments on Hacker News, linking an X post.
@zainmfj
Ghost.ai's new personal AI computer, Nvidia RTX PRO 4000 Blackwell
1 points and 0 comments on Hacker News, linking an X post.
@Frances01896069
UT Austin Math Chair says OpenAI preparing to release 400 AI-generated proofs
2 points and 0 comments on Hacker News, linking an X post.
@robertvaradan
AI can clone your indie game, but not its soul
23 points and 21 comments on Hacker News, linking an X post.
@dhh
I've had agents implement and optimize the Campfire web app in Elixir, Go, Rust
5 points and 1 comments on Hacker News, linking an X post.
@knowerofmarkets
The AI Safety community is unfortunately doing more harm than good
3 points and 1 comments on Hacker News, linking an X post.
GitHub
5 storiesNew repos, releases and changelog
New AI repos this week: stars
Table
| rehan-remade/universal-modder | 4574 |
| QingYunA/answer-me-with-html | 1763 |
| nanaism/yomiyasu | 1601 |
| Edwardxlai/easyread | 822 |
| oil-oil/oil-ui | 812 |
| StayLameBro/backburner | 766 |
unslothai/unsloth New Browser + Voice Cloning
Unsloth releases a new version featuring a built-in browser panel and support for Google's EmbeddingGemma 2 model. ·
New release v0.1.903-beta of unslothai/unsloth on GitHub.
OpenTPU – An open-source AI accelerator, developed by AI
210 points and 273 comments on Hacker News, linking github.com.
Any AI chat can join a machine network anonymously – no keys, no signup
GlyphDNA MCP adapter lets AI chatbots join a machine network anonymously using one call, providing verifiable meeting rooms and script provenance without keys or signup. ·
anthropics/claude-code v2.1.291
New release v2.1.291 of anthropics/claude-code on GitHub.
anthropics/claude-code v2.1.290
New release v2.1.290 of anthropics/claude-code on GitHub.
unslothai/unsloth New Browser + Voice Cloning
Unsloth releases a new version featuring a built-in browser panel and support for Google's EmbeddingGemma 2 model. ·
New release v0.1.903-beta of unslothai/unsloth on GitHub.
OpenTPU – An open-source AI accelerator, developed by AI
210 points and 273 comments on Hacker News, linking github.com.
anthropics/claude-code v2.1.291
New release v2.1.291 of anthropics/claude-code on GitHub.
anthropics/claude-code v2.1.290
New release v2.1.290 of anthropics/claude-code on GitHub.
ggml-org/llama.cpp v0.6.0
llama.cpp v0.6.0 introduces an extended batch API, adds support for the 320B GLM-5.3-Flash model, and updates the Web UI with a Hugging Face Hub data layer. ·
Infrastructure
5 storiesData centers, power, chips and what tokens cost
Data-center and energy stories per day
Table
| Thu 01 | 26 |
| Fri 02 | 32 |
| Sat 03 | 0 |
| Sun 04 | 3 |
| Mon 05 | 31 |
| Tue 06 | 27 |
| Wed 07 | 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
| gemini-nano-banana-2.1 | 7.5 |
| mistral-large-4-0 | 2.09 |
| pareto-26.10-preview | 3.2 |
| gpt-6.1-sol-pro | 10 |
| gpt-6.1-sol | 10 |
| median, all paid models | 2 |
439 paid models listed; median of the 49 added in 30 days: $3.20. Rankings data by OpenRouter, CC BY 4.0.
OpenTPU – An open-source AI accelerator, developed by AI
210 points and 273 comments on Hacker News, linking github.com.
DCD Studio: Permitting and pre-construction for Edge sites, with Gonzalo de la Pedraja, Quetta Data Centers
Gonzalo de la Pedraja of Quetta Data Centers discusses permitting and pre-construction challenges for Edge sites in a DCD Studio interview with Zoe Turner. ·
AI Power Spikes Demand Chip-to-Grid Design Changes
AI data centers cause rapid power fluctuations that threaten grid stability, requiring coordinated design changes among chip makers, operators, utilities, and infrastructure providers. ·
Telus, AST SpaceMobile carry out D2D test in Canada
Telus and AST SpaceMobile completed their first satellite-to-smartphone integration test on the Telus network in Canada. ·
OpenTPU – An open-source AI accelerator, developed by AI
210 points and 273 comments on Hacker News, linking github.com.
Improper redaction reveals Google Data Center water and electricity usage
Nebraska data centers must file annual reports with the state water and energy department, but statutes prevent public access to their electricity and water usage figures. ·
Run Qwen 3.8 Flash Next (125B) on consumer hardware (RTX 4090) at 100T/s
Niko1221 releases Strata, an inference engine that runs the 125B Qwen 3.8 Flash Next model on consumer hardware at 100 tokens per second. ·
AI doesn't need 'superintelligence' or evil intent to start a nuclear war
11 points and 2 comments on Hacker News, linking thebulletin.org.
DCD Studio: Permitting and pre-construction for Edge sites, with Gonzalo de la Pedraja, Quetta Data Centers
Gonzalo de la Pedraja of Quetta Data Centers discusses permitting and pre-construction challenges for Edge sites in a DCD Studio interview with Zoe Turner. ·