News
The raw AI news stream — everything we ingest across RSS, Hacker News, arXiv, GitHub, Bluesky, YouTube and more, newest first.
| Source | Age | Title | Summary | Status |
|---|---|---|---|---|
arxiv arxiv-ai | 1mo | Iris-3B: Going Beyond the Latent with Pixel-Space Diffusion Training, Conversion and Fine-Tuning🔧 FLUX | Pixel-space diffusion models avoid the lossy VAE of latent models, which suggests an advantage on downstream tasks where fine-grained… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Das KI-Unternehmen hat ein internes KI-Modell auf tausende Probleme aus der Mathematik angesetzt. Für ein Zehntel davon hat es relevante Lö… | Das KI-Unternehmen hat ein internes KI-Modell auf tausende Probleme aus der Mathematik angesetzt. Für ein Zehntel davon hat es relevan… | ↗Not briefed |
bluesky bluesky-ai | 1mo | openai should release future math results as embeddings | openai should release future math results as embeddings | ↗Not briefed |
arxiv arxiv-ai | 1mo | Arctic Questions, Missing Answers: A Dataset and Benchmark for LLM Abstention in Arctic Science🔧 ChatGPT | Large language models (LLMs) should abstain from scientific multiple-choice questions when no option is valid, but frequent abstention… | ↗Not briefed |
rss latent-space | 1mo | [AINews] Quasi-Riemann-Hypothesis: OpenAI publishes 722 math papers solving 90 of the top 500 open math problems; “the most significant moment” in >100 years of mathematics | Our head hurts. | ↗Not briefed |
hn hn-ai | 1mo | Show HN: KORA Doctor – find LLM calls your agent may not have needed | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | So what do you do if you're a lab that has a model that can solve problems at that level? For a start you pump the gas on it a bit, to see … | So what do you do if you're a lab that has a model that can solve problems at that level? For a start you pump the gas on it a bit, to… | ↗Not briefed |
bluesky bluesky-ai | 1mo | OpenAI releases new cookbook “To Serve Man” | OpenAI releases new cookbook “To Serve Man” | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Snug, a local-only bookmark export, import and backup extension | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | If you don’t take the time to write the words—if you have an LLM build a composite document for you—your end product is not worth reading. … | If you don’t take the time to write the words—if you have an LLM build a composite document for you—your end product is not worth read… | ↗Not briefed |
bluesky bluesky-ai | 1mo | OpenAI solves hardest problems in political science; the optimal form of governance is “direct rule by the House of Windsor”, very surprisi… | OpenAI solves hardest problems in political science; the optimal form of governance is “direct rule by the House of Windsor”, very sur… | ↗Not briefed |
bluesky bluesky-ai | 1mo | "The problem is that companies like OpenAI and Anthropic are rushing to IPO to raise billions of dollars for their founders." @tobywalsh.bs… | "The problem is that companies like OpenAI and Anthropic are rushing to IPO to raise billions of dollars for their founders." @tobywal… | ↗Not briefed |
bluesky bluesky-ai | 1mo | I spend a surprising amount of time at work explaining to non engineers all the reasons we should use them instead of a million dollars of … | I spend a surprising amount of time at work explaining to non engineers all the reasons we should use them instead of a million dollar… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Can OpenAI just die already? | Can OpenAI just die already? | ↗Not briefed |
arxiv arxiv-ai | 1mo | Mixture of Layers: Dynamic Layer Routing for Visual Reasoning | Pre-trained vision encoders contain layer-wise visual representations that differ in spatial granularity, semantic abstraction, and se… | ↗Not briefed |
hn hn-ai | 1mo | Crunchbase Data Shows AI's Most Active Startups Are Becoming Serial Acquirers | — | ↗Not briefed |
rss simon-willison | 1mo | Quoting Jake Boggan | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | 生成系ＡＩを使ってる学生ＬＬＭが創発的に出力作っていると信じている事に背筋が寒くなる。 | 生成系ＡＩを使ってる学生ＬＬＭが創発的に出力作っていると信じている事に背筋が寒くなる。 | ↗Not briefed |
hn hn-ai | 1mo | The reason Anthropic must IPO now [video] | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | An extended deep energy method for thermo-mechanical crack propagation | Thermo-mechanical fracture couples transient heat conduction on a cracked domain with a crack that grows as the temperature and the di… | ↗Not briefed |
bluesky bluesky-ai | 1mo | this would be like me sending an LLM to play a ttrpg with my friends (mind you they might enjoy the experience better that way) | this would be like me sending an LLM to play a ttrpg with my friends (mind you they might enjoy the experience better that way) | ↗Not briefed |
bluesky bluesky-ai | 1mo | me if I were a mathematician right now: "I ain't reading all that. I'm happy for u tho." gizmodo.com/openai-dumps... | me if I were a mathematician right now: "I ain't reading all that. I'm happy for u tho." gizmodo.com/openai-dumps... | ↗Not briefed |
bluesky bluesky-ai | 1mo | 一般人がLLMをおふざけや宿題に使う一方で，一部のロボット工学者はロボットを事前プログラミングの制約から解放する方法としてLLMに注目している。 別冊日経サイエンス289 『SFを科学する 士郎正宗を作るには』 www.nikkei-science.com/sci_book… | 一般人がLLMをおふざけや宿題に使う一方で，一部のロボット工学者はロボットを事前プログラミングの制約から解放する方法としてLLMに注目している。 別冊日経サイエンス289 『SFを科学する 士郎正宗を作るには』 www.nikkei-science.com/sci… | ↗Not briefed |
bluesky bluesky-ai | 1mo | A verified “Pirate Bay” for open LLMs, with torrent-ready model weights, hosted mirrors, license clarity, source provenance, and SHA-256 ha… | A verified “Pirate Bay” for open LLMs, with torrent-ready model weights, hosted mirrors, license clarity, source provenance, and SHA-2… | ↗Not briefed |
arxiv arxiv-ai | 1mo | RSI-Forge: From Research Papers to Environments for Recursive Self-Improvement | Environments are the foundation of recursive self-improvement: they provide the problems agents work on and the feedback used to evalu… | ↗Not briefed |
bluesky bluesky-ai | 1mo | I asked Google what would be in the readme file | I asked Google what would be in the readme file | ↗Not briefed |
arxiv arxiv-ai | 1mo | What a Reporting Convention Hides: A Matched-Budget Audit of Quantum Natural Gradient with an Exactly Computed Metric | Several published comparisons of variational quantum optimizers time only runs that reach a target loss, or read the verdict at a sing… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Democratizing MoE inference on commodity GPUs with CoMoE | Deploying Mixture-of-Experts (MoE) models relies heavily on Expert Parallelism, which generates intense inter-GPU communication. Conse… | ↗Not briefed |
bluesky bluesky-ai | 1mo | youtu.be/uJcPjrEgppk [Deepseek & Huawei Made Local AI Run Without NVIDIA's CUDA (And Why it's Hard)] | youtu.be/uJcPjrEgppk [Deepseek & Huawei Made Local AI Run Without NVIDIA's CUDA (And Why it's Hard)] | ↗Not briefed |
arxiv arxiv-ai | 1mo | Efficient Reasoning with Flow Language Models | Flow Language Models (FLMs) have emerged as a continuous-state alternative to discrete diffusion language models, yet the role of thei… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Self-Consuming Generative Models with Co-Evolving Human Preferences | Generative models are increasingly trained in self-consuming iterative loops, where users curate preferred samples from model-generate… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Sony has filed a patent that would take data generated by the player and use it to train NPCs behaviors. https://bit.ly/4ziZGQP | Sony has filed a patent that would take data generated by the player and use it to train NPCs behaviors. https://bit.ly/4ziZGQP | ↗Not briefed |
arxiv arxiv-ai | 1mo | Finding the Right Balance: Relevance and Diversity in LLM Retrieval | Retrieval diversification is widely available in retrieval-augmented generation (RAG) frameworks, yet prior studies disagree on whethe… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Shared Geometry As A Rosetta Stone: Cross-Modal Alignment Without Paired Data🔧 Modal | Multimodal representations enable zero-shot classification and retrieval, but aligning independently trained models usually requires l… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Stability and Diversity of Networked Self-Consuming Generative Ecosystems | The widespread deployment of generative AI has made it increasingly difficult to distinguish synthetic content from real data. Consequ… | ↗Not briefed |
bluesky bluesky-ai | 1mo | I literally said this to my mom like three weeks ago and her responses was like "well yes of course these companies want to create slaves" … | I literally said this to my mom like three weeks ago and her responses was like "well yes of course these companies want to create sla… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Don't make me tap the sign. Just because a company is private does not make it a startup. OpenAI has raised $180 billion or so and has abou… | Don't make me tap the sign. Just because a company is private does not make it a startup. OpenAI has raised $180 billion or so and has… | ↗Not briefed |
hn hn-ai | 1mo | Claude subscription plan provides more value than OpenAI's, study says | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Noise, Denoise, Correct: MCMC Posterior Sampling with Diffusion Priors in Three Steps | Pretrained diffusion models are powerful priors for inverse problems, but posterior sampling under nonlinear, non-differentiable forwa… | ↗Not briefed |
bluesky bluesky-ai | 1mo | LLMに試しに脚注を書かせようとしましたが，まだHal Tasaki的なセンスを感じないですね．ボルツマンが出てきたタイミングで，「ボルツマンは，実は，ブルックナーにピアノを習っていた．ベートーヴェンの交響曲をピアノで弾くのが好きだった．」みたいな脚注を書くセンス． | LLMに試しに脚注を書かせようとしましたが，まだHal Tasaki的なセンスを感じないですね．ボルツマンが出てきたタイミングで，「ボルツマンは，実は，ブルックナーにピアノを習っていた．ベートーヴェンの交響曲をピアノで弾くのが好きだった．」みたいな脚注を書くセンス． | ↗Not briefed |
arxiv arxiv-ai | 1mo | Let the Library Speak: Self-Advertised Method Selection for Formal Proving | LLM-based formal provers can retrieve relevant lemmas and prior proofs, but relevance alone does not say whether a mathematical method… | ↗Not briefed |
arxiv arxiv-ai | 1mo | TutorLoop: Regulating Student Learning Behaviors via Sensor-in-the-Loop Generative Feedback | We present TutorLoop, a sensor-in-the-loop system that regulates student learning behaviors by delivering adaptive feedback based on r… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Git extension for controlling Git worktree mess | we built a git extension and we're inviting the community to help us build a better way than just using git worktree. | ↗Not briefed |
bluesky bluesky-ai | 1mo | you don’t like ai? have you tried it? how are you supposed to have an informed opinion unless you try it? oh you have? what model? that’s a… | you don’t like ai? have you tried it? how are you supposed to have an informed opinion unless you try it? oh you have? what model? tha… | ↗Not briefed |
rss techmeme | 1mo | AI startup acquisitions by other AI startups hit 195 through Sept. 29, up 14% from all of 2025, but number of buyers grew just 2%; OpenAI was the most active (Mary Ann Azevedo/Crunchbase News) | Mary Ann Azevedo / Crunchbase News: AI startup acquisitions by other AI startups hit 195 through Sept. 29, up 14% from all of 2025, bu… | ↗Not briefed |
arxiv arxiv-ai | 1mo | ARCS: Towards Precise Text-to-SQL via Structured Disambiguation | As text-to-SQL systems move beyond demonstrations toward real-world deployment, ambiguity in user questions becomes a primary source o… | ↗Not briefed |
bluesky bluesky-ai | 1mo | I’m not sure about a lot of things in this world, but I’m sure of this: if we start a witch hunt, we lose. Please shut this British guy up… | I’m not sure about a lot of things in this world, but I’m sure of this: if we start a witch hunt, we lose. Please shut this British g… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Wow. Plaintiffs lawyers didn't even try to defend that "key" report they had questionably cooked up to fill in the gaps to attack fair use … | Wow. Plaintiffs lawyers didn't even try to defend that "key" report they had questionably cooked up to fill in the gaps to attack fair… | ↗Not briefed |
bluesky bluesky-ai | 1mo | will anyone besides OpenAi even be able to train on these results in this state? will math knowledge cutoffs have these in them? well proba… | will anyone besides OpenAi even be able to train on these results in this state? will math knowledge cutoffs have these in them? well… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Sorry “with guesslation” I was trying to say This is before LLM/Ai stuff - so basically either DeepL or Google Translate. There was a seco…🔧 DeepL | Sorry “with guesslation” I was trying to say This is before LLM/Ai stuff - so basically either DeepL or Google Translate. There was a… | ↗Not briefed |