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 |
|---|---|---|---|---|
bluesky bluesky-ai | 1mo | これは百も承知でLLMと現在のweb環境を批判指摘してる。これに対してくろこは、そうですね。だから飛行機は落ちたし、これからも落ちると思っていますし、そう返します | これは百も承知でLLMと現在のweb環境を批判指摘してる。これに対してくろこは、そうですね。だから飛行機は落ちたし、これからも落ちると思っていますし、そう返します | ↗Not briefed |
hn hn-ai | 1mo | Narwhal: Adaptive disaggregated inference in a role free fleet | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Beyond Imitation: A Framework and Benchmark for LLM-Assisted Peer Review | The rapid growth of scientific publishing has strained peer review, particularly in machine learning, raising concerns about declining… | ↗Not briefed |
bluesky bluesky-ai | 1mo | 'OpenAI, what's a mommy?' | 'OpenAI, what's a mommy?' | ↗Not briefed |
bluesky bluesky-ai | 1mo | Terrence Tao shared the parody headline over on mathstodon: "OpenAI Releases the Final Ten Minutes of 500 Previously Unreleased Films, Ushe… | Terrence Tao shared the parody headline over on mathstodon: "OpenAI Releases the Final Ten Minutes of 500 Previously Unreleased Films,… | ↗Not briefed |
bluesky bluesky-ai | 1mo | just did a twitter thread on this recent geometry dash level by a new creator that was ai generated, thought you guys should know this stuf… | just did a twitter thread on this recent geometry dash level by a new creator that was ai generated, thought you guys should know this… | ↗Not briefed |
arxiv arxiv-ai | 1mo | DaCe-DT: Data-Centric Offline Multi-Task Reinforcement Learning via Adaptive Prompts and Trajectory Correction for Heterogeneous Tasks | Offline multi-task reinforcement learning (Offline MTRL) heavily depends on the quality and distribution of pre-collected data. Howeve… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Agentc: <1MB static, no Libc, minimal, embeddable agent with Rust/C API | Sharing an agent that I built since most of the agents that's nice to work with, take hefty deps like node or is provider depende… | ↗Not briefed |
bluesky bluesky-ai | 1mo | As Canberra tries to thread the needle on AI, it's becoming gradually more enmeshed with companies like OpenAI and Anthropic, writes Ben El… | As Canberra tries to thread the needle on AI, it's becoming gradually more enmeshed with companies like OpenAI and Anthropic, writes B… | ↗Not briefed |
bluesky bluesky-ai | 1mo | OpenAI, next month: "PI'nglui mglw'navier Cthulhu R'lyeh wgah'nagl fhtagnpi/2" | OpenAI, next month: "PI'nglui mglw'navier Cthulhu R'lyeh wgah'nagl fhtagnpi/2" | ↗Not briefed |
rss marktechpost | 1mo | What Happens When a Trusted Model Repo Changes? Unsloth Studio Re-Checks Before It Runs | Unsloth's October 6 security overview explains how Studio checks code, weights, packages and tools before anything runs. Custom model… | ↗Not briefed |
bluesky bluesky-ai | 1mo | It’s occurring to me that Anthropic did not advertise the existence of this model because they’re good guys who want China to keep releasin… | It’s occurring to me that Anthropic did not advertise the existence of this model because they’re good guys who want China to keep rel… | ↗Not briefed |
bluesky bluesky-ai | 1mo | What if OpenAI figures out that Jewish People were right? | What if OpenAI figures out that Jewish People were right? | ↗Not briefed |
bluesky bluesky-ai | 1mo | 「Claude Haiku 5.5」が登場、GPT-6 Lunaより高性能で前モデルより9割引きの安価なモデル https://gigazine.net/news/20261008-claude-haiku-5-5/ | 「Claude Haiku 5.5」が登場、GPT-6 Lunaより高性能で前モデルより9割引きの安価なモデル https://gigazine.net/news/20261008-claude-haiku-5-5/ | ↗Not briefed |
bluesky bluesky-ai | 1mo | A picture of OpenAI solving math problems will never look this cool. | A picture of OpenAI solving math problems will never look this cool. | ↗Not briefed |
bluesky bluesky-ai | 1mo | The only reason why I haven't deactivated my twitter account is to keep others from getting a hold of my username and using it. But the to… | The only reason why I haven't deactivated my twitter account is to keep others from getting a hold of my username and using it. But t… | ↗Not briefed |
arxiv arxiv-ai | 1mo | A General $\widetildeΩ(\sqrt{T γ_T})$ Lower Bound for Kernel Bandits | The kernel bandit problem consists of sequentially optimizing an unknown function with noisy feedback, where the function has bounded… | ↗Not briefed |
bluesky bluesky-ai | 1mo | if you take modding/programming help, advice, or guidance from me, DO NOT EVER under ANY circumstance feed it to an LLM to do the work for … | if you take modding/programming help, advice, or guidance from me, DO NOT EVER under ANY circumstance feed it to an LLM to do the work… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Fascinating to watch artists go from drawing in their own style in an iterative process with the commissioner to quickly churning out samey… | Fascinating to watch artists go from drawing in their own style in an iterative process with the commissioner to quickly churning out… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Did they try Deepseek tho? | Did they try Deepseek tho? | ↗Not briefed |
arxiv arxiv-ai | 1mo | Stability-Plasticity Balance via Singular-Vector Selection in LLM Continual Learning | Domain-specific continual adaptation of LLMs risks catastrophic forgetting, creating a fundamental tension between acquiring new capab… | ↗Not briefed |
bluesky bluesky-ai | 1mo | or just go all in on deepseek | or just go all in on deepseek | ↗Not briefed |
arxiv arxiv-ai | 1mo | Optimally Pacing Budget Spending and Learning | We establish near-optimal regret bounds for budget-constrained online learning against arbitrary classes of budget-pacing experts in t… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Clinician use of language models diverges from how the models are evaluated | Large language model (LLM) assistants are being deployed to clinicians across health systems, and judgments about their readiness rest… | ↗Not briefed |
bluesky bluesky-ai | 1mo | RIP Margret Hamilton aka rope mother. Thanks for getting us to the moon. And she did it all without an LLM or even stack overflow. A true … | RIP Margret Hamilton aka rope mother. Thanks for getting us to the moon. And she did it all without an LLM or even stack overflow. A… | ↗Not briefed |
bluesky bluesky-ai | 1mo | I wish OpenAI put more effort into making these math results human-understandable, if nothing else it would be good practice for when AI st… | I wish OpenAI put more effort into making these math results human-understandable, if nothing else it would be good practice for when… | ↗Not briefed |
arxiv arxiv-ai | 1mo | CityDeploy-Bench: Benchmarking Physics-Grounded Spatial Set Planning for Multi-Transmitter Network Deployment | Automating city-scale wireless deployment remains challenging under complex urban propagation and network-wide interference. We introd… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Measuring and Mitigating Solution Mode Collapse in RLVR | A language model (LM) can usually answer the same question in more than one way, but reinforcement learning with verifiable rewards (R… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Emergent Inverse-Depth Scaling From Nonlinearity In Attention | Scaling laws describe power-law improvements in model performance with dataset size and parameter count, yet their underlying mechanis… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Anthropic will be 'most ridiculous IPO' of year, analyst says www.cnbc.com/id/108374300... | Anthropic will be 'most ridiculous IPO' of year, analyst says www.cnbc.com/id/108374300... | ↗Not briefed |
hn hn-ai | 1mo | NotProton: Steam Play for macOS | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | It takes about 250 documents to plant a backdoor in an AI model. Your chats train the models unless you find the switch; business customers… | It takes about 250 documents to plant a backdoor in an AI model. Your chats train the models unless you find the switch; business cust… | ↗Not briefed |
hn hn-ai | 1mo | Agent in a Bottle: Can LLM Agents Turn Their Capabilities into Cheap Artifacts? | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | Harvard just dropped a study AI productivity in software engineering and it found the average increase in productivity from using Claude co… | Harvard just dropped a study AI productivity in software engineering and it found the average increase in productivity from using Clau… | ↗Not briefed |
arxiv arxiv-ai | 1mo | AffordDrive3D: Affordance-Aware World-Action Modeling with Spatial Understanding | World-action models have recently improved autonomous driving by jointly learning future scene prediction and trajectory generation. M… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Learning What to Trust in Multimodal Learning under Noisy Supervision | Multimodal classification processes and relates information from multiple modalities to achieve more accurate predictions. However, ex… | ↗Not briefed |
bluesky bluesky-ai | 1mo | The fun part about people having "poor memories" about inter-factional beefs (beeves?) is that I have the Twitter archives and a shitload o… | The fun part about people having "poor memories" about inter-factional beefs (beeves?) is that I have the Twitter archives and a shitl… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Given that every other feature that promises *not* to do the work will do it if you ask, curious to know how soon someone will sue OpenAI w… | Given that every other feature that promises *not* to do the work will do it if you ask, curious to know how soon someone will sue Ope… | ↗Not briefed |
bluesky bluesky-ai | 1mo | AI whistleblowers testify about the technology’s dangers. “Jacob Coxon, the former Anthropic researcher who sparked industry chaos by warni… | AI whistleblowers testify about the technology’s dangers. “Jacob Coxon, the former Anthropic researcher who sparked industry chaos by… | ↗Not briefed |
rss techmeme | 1mo | Keyu Tian, a former ByteDance intern, raised ~$30M from 5Y and IDG for his unnamed AI lab that focuses on building world models, at a $200M post-money valuation (Bloomberg) | Bloomberg: Keyu Tian, a former ByteDance intern, raised ~$30M from 5Y and IDG for his unnamed AI lab that focuses on building world mo… | ↗Not briefed |
arxiv arxiv-ai | 1mo | AgentHorizon: Evaluating Agentic Judges for Long-Horizon Computer-Use Tasks🔧 GPT-5 | Computer-use agents are capable of completing complex tasks, increasing the use of automatic judges to determine success, either for t… | ↗Not briefed |
bluesky bluesky-ai | 1mo | For some reason everyone is dogging on you when you have an entirely valid and correct point to being up. Why are we keeping huge 360 (noto… | For some reason everyone is dogging on you when you have an entirely valid and correct point to being up. Why are we keeping huge 360… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Another day, another apocalypse averted by not believing the hype. pivot-to-ai.com/2026/10/07/m... | Another day, another apocalypse averted by not believing the hype. pivot-to-ai.com/2026/10/07/m... | ↗Not briefed |
bluesky bluesky-ai | 1mo | Anthropic, OpenAI, Facebook, Xai, none of them have a Margaret Hamilton. | Anthropic, OpenAI, Facebook, Xai, none of them have a Margaret Hamilton. | ↗Not briefed |
bluesky bluesky-ai | 1mo | in about three years the c-suite getting their minds liquefacted by daily llm exposure will be talking to eachother like that. you could ex… | in about three years the c-suite getting their minds liquefacted by daily llm exposure will be talking to eachother like that. you cou… | ↗Not briefed |
hn hn-ai | 1mo | OpenAI's human rights lead: What military AI could do 'keeps me up at night' | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | 「日販が日本の書籍を大量にanthropicに売った話」よりも、 毎日新聞がGoogleに過去ニュースの縮刷版をデータベース化させる契約をしてごっそり大規模学習をさせた話のほうがメガトン級にやばいんだけど 新聞業界はその辺の話は庇うんだよね… | 「日販が日本の書籍を大量にanthropicに売った話」よりも、 毎日新聞がGoogleに過去ニュースの縮刷版をデータベース化させる契約をしてごっそり大規模学習をさせた話のほうがメガトン級にやばいんだけど 新聞業界はその辺の話は庇うんだよね… | ↗Not briefed |
bluesky bluesky-ai | 1mo | ok maybe math is cooked | ok maybe math is cooked | ↗Not briefed |
rss techmeme | 1mo | Mecka, which collects human motion data to train humanoid robots, raised a $60M Series B led by Sequoia, with participation from Nvidia, M12, and others (Marina Temkin/TechCrunch) | Marina Temkin / TechCrunch: Mecka, which collects human motion data to train humanoid robots, raised a $60M Series B led by Sequoia, w… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Then they could work with mathematicians to convert them into papers as Anthropic recently did | Then they could work with mathematicians to convert them into papers as Anthropic recently did | ↗Not briefed |