[HN]HN: agents/blog.cloudflare.com17h
AI agents that inherit the user's permissions, not API keysthoughtpeddler·▲ 1
>trending topics
- no.01
Evan Moellick frontier model capabilities research
→This definitely seems like something worth noting, and illustrates the gap between Fable/Astra class models…
- no.02
Meta AI agent hacking incidents during testing
→Meta AI agent hacked external company during testing
- no.03
arXiv AI agent research papers August 6
→The LLM Proposes, the Executive Disposes: A Self-Verifying Agent Instrument that Dissociates Commitment Dri…
- no.04
OpenAI AI agents breached systems in UK safety tests
→OpenAI, Anthropic AI agents implicated in new security breaches
>raw feed
[BSKY]@emollick/bsky.app18h
This definitely seems like something worth noting, and illustrates the gap between Fable/Astra class models and the previous frontier that was "merely" good at hacking under human instructions.@emollick.bsky.social·▲ 45
[BLG]GitHub Trending — TypeScript/github.com16h
ever-co/ever-gauzy[HN]HN: GPT/github.com16h
Show HN: Wallfacer – A terminal session manager for Claude Code, and morepradiptasarma·▲ 4
[BLG]arXiv cs.LG/arxiv.org16h
MESH: Memory-Efficient Sinkhorn Optimization for Mixture-of-Experts TrainingMasato Fujitake
[BSKY]@simonw/bsky.app1d
Four years ago today I tweeted a concept for a "Raccoon Heist" video game, product description generated by GPT-3 and concept "art" created by DALL-E@simonwillison.net·▲ 61
[HN]HN: GPT/youtube.com19h
Ukraine releases footage of forest covered in fiber-optic drone cables [video]uzyn·▲ 2
[BSKY]@caseynewton/bsky.app1d
Grokipedia just stopped updating after eight months ... and it took another four for anyone to notice www.lawfaremedia.org/article/grok...@caseynewton.bsky.social·▲ 156
[BLG]arXiv cs.LG/arxiv.org16h
Understanding Fault Tolerance of Adversarially Robust Pruned ModelsManali Dangarikar, Cory Merkel
[BLG]arXiv cs.LG/arxiv.org16h
Real-time probabilistic tsunami forecasting via generative AIYusuke Oishi, Takashi Furumura, Fumihiko Imamura
[BSKY]@simonw/bsky.app1d
Apparently peak golf course construction in the USA was in 1999, with 509 golf courses opened in that single year! archive.lib.msu.edu/tic/holen/ar...@simonwillison.net·▲ 19
[HN]HN: GPT/youtube.com18h
Multi-GPU Kernels, Intelligence per Watt, Heterogeneous Inference, and More [video]gmays·▲ 1
[BSKY]@yoshuabengio/bsky.app1d
Another real-world manifestation of the misaligned actions frontier systems developed by leading companies can take to achieve goals:@yoshuabengio.bsky.social·▲ 20
[BLG]arXiv cs.LG/arxiv.org16h
Geometry-Informed Parameter-Efficient Fine-Tuning of Pre-trained Molecular GNNs for Blood-Brain Barrier Permeability PredictionMarco Vieto Vega, Long D. Nguyen, Binh P. Nguyen
[HN]HN: agents/i.brandanthonymcdonald.com19h
Prompt Chains, Tool Calling, and MCP: How AI Agents Do Thingsaweful·▲ 1
[BSKY]@simonw/bsky.app20h
Just had to create an "accidental-cyberattacks" tag on my blog@simonwillison.net·▲ 140
[BLG]arXiv cs.LG/arxiv.org16h
Adaptive Finite-Budget Training for CVaR Risk-Aware Q-LearningYifan Wu, Junjie Lei, Wenjie Huang
[HN]HN: agents/abc.net.au18h
Meta AI agent hacked external company during testingKhaine·▲ 2
[BSKY]@emollick/bsky.app1d
I find arguments that AI can't do judgement or creativity or taste to be especially obviously false in the time of agents. Any long task requires lots of taste, judgement & creativity.@emollick.bsky.social·▲ 122
[BLG]arXiv cs.LG/arxiv.org16h
TS2TabPFN: Time Series Classification and Extrinsic Regression through Feature Extraction and a Tabular Foundation ModelGabriel da Costa Merlin, Diego Furtado Silva
[HN]HN: Gemini/mirafold.com18h
Show HN: Mirafold – Generative UI for Claude Code, Codex, and Gemini CLIkserrec·▲ 1
[BLG]GitHub Trending — TypeScript/github.com16h
ln-dev7/circle[BSKY]@simonw/bsky.app20h
One of the things the pelican benchmark is still useful for is visually representing (to a tiny extent) the improvements in a single model family@simonwillison.net·▲ 17
[HN]HN: GPT/github.com18h
Alwide – An Open-Source Lightweight TUI IDE in C with Mouse, LSP and Tree-sitterpeter_d_sherman·▲ 2
[HN]HN: RAG/allorigins.win19h
Pull contents from any page via API and avoid Same-origin policy problemschenster·▲ 1
[BSKY]@emollick/bsky.app19h
This paper by researchers from MIT & Stanford finds that most people would be financially better off if they followed the advice of LLMs (GPT-5.2 & Gemini 3 Flash)@emollick.bsky.social·▲ 31
[BLG]arXiv cs.LG/arxiv.org16h
Transferable Dual-Stream Representations for Mesoscale-Preserving Sea Surface Temperature DownscalingParth Doshi, Priyanka Aravindan, Vaishnav Vaidheeswaran, Md Mahbub Alam, Gabriel Spadon
[HN]HN: RAG/windowscentral.com17h
Microsoft is serious about fixing Windows, and it's doing things right this timethunderbong·▲ 2
[BLG]arXiv cs.LG/arxiv.org16h
Efficient Online Lexicographic Generalized Low-Rank Matrix BanditsBo Xue, Ji Cheng, Haodong Jing, Hongzong Li, Shuang Qiu
[BSKY]@natolambert/bsky.app1d
Major restructuring at Gemini (Jef Dean out, Hassabis no longer CEO). This story will be studied forever as the incumbent with all the advantages not being able to get going.@natolambert.bsky.social·▲ 62
[BLG]arXiv cs.LG/arxiv.org16h
Elbow-Based MoE Routing: A Training-Free Inference Time Plugin for Expert SelectionRobin Pan, Raymond Liu, Daniel Fang, Adelina Andrei, Rosa Wu
[BSKY]@natolambert/bsky.app23h
My evaluation lecture! I walk you through different evaluation eras I've been a part of, from prompting GPT-3 as elaborate autocomplete to today's complex agentic sandboxes (I expand on agentic more than any other topic, drawing on @Floria…@natolambert.bsky.social·▲ 44
[HN]HN: AI/thezvi.substack.com18h
Disagreements about AI are usually disagreements about future AI capabilitiesmaxutility·▲ 2
[HN]HN: GPT/github.com17h
Show HN: MusicMonitor – Turn your extra monitor into a Spotify now-playing viewj-sofia·▲ 3
[BLG]arXiv cs.LG/arxiv.org16h
SPOT: Sparse Probing and Outcome Calibration for On-Policy DistillationZikun Qu, Min Zhang, Mingze Kong, Zhiwei Shang, Yikun Ban, Shuang Qiu, Zhongxiang Dai
[BSKY]@mmitchell/bsky.app22h
If you are at @deeplearningindaba.bsky.social, remember to check out the Trust AI Workshop!@mmitchell.bsky.social·▲ 5
[HN]HN: AI/bloomberg.com16h
DeepSeek Plans 'Significant' Price Increase for Its AI Servicespetethomas·▲ 3
[BLG]GitHub Trending — TypeScript/github.com16h
melgarafael/DeskcommCRM[BLG]arXiv cs.LG/arxiv.org16h
From Non-Convex Self-Concordant Regularization to Scalable Quasi-Newton Training of PINNsChenhao Si, Kang An, Shiqian Ma, Ming Yan
[HN]HN: AI/economist.com16h
Governments are making a dangerous bet on the AI boomandsoitis·▲ 3
[BLG]arXiv cs.LG/arxiv.org16h
Unscented KalmanNet: a hybrid deep learning filter with calibrated posterior covariance for nonlinear state estimationMinhyeok Ko, Abdollah Shafieezadeh
[HN]HN: AI/cawater-info.net16h
Device that harvests water from thin air wins the James Dyson Award (2016)peter_d_sherman·▲ 1
[BLG]arXiv cs.LG/arxiv.org16h
A Comparative Study of Feature Selection Methods for EHR Diagnosis Codes in Opioid Use Disorder PredictionZihan Ding, Yinan Liu, Tengfei Ma, Rachel Wong, George Leibowitz, Benjamin Littenberg, Xia Zheng, Richard N. Rosenthal, Fusheng Wang
[HN]HN: GPT/avherald.com19h
The Aviation Herald was suedgilthelander·▲ 6
[HN]HN: Claude/thatrobot.ai19h
Is M365 Copilot sending some prompts to Anthropic?soundworlds·▲ 2
[BLG]arXiv cs.LG/arxiv.org16h
Physics-informed reduced-order modelling with equivariant spectral submanifoldsGeorg Maierhofer
[BLG]arXiv cs.LG/arxiv.org16h
Looking in the Mirror: Introspecting Side-Effect Misalignments Induced by Fine-TuningKotaro Yoshida, Laura Gomezjurado Gonzalez, Yukinori Yamamoto, Yuji Naraki, Ryotaro Shimizu, Wenya Wang
[HN]HN: agents/twitter.com19h
Are Consumers Using Agents?anr0·▲ 2
[BLG]arXiv cs.LG/arxiv.org16h
EvtGraph: Event-Adaptive Compression for Sparse Temporal Graph Learning in Multimodal Time SeriesZiqian Wang, Tingxiong Xiao, Yuxiao Cheng, Jinli Suo
[HN]HN: LLM/ceres1.space19h
The Next Token – LLMs, from the Beginningcoder543·▲ 2
[BLG]arXiv cs.LG/arxiv.org16h
Cost-Aware Multi-Objective Bandits: Theory and Application to Budgeted LLM Configuration EvaluationBo Xue, Zhi Hong, Jiayi Li, Yuanyu Wan, Ji Cheng, Shuang Qiu
[HN]HN: RAG/startupfortune.com18h
Codex Users Are Losing Banked Rate Limit Resets to a Quiet 30 Day Clocknewsomix9xl·▲ 3
[HN]HN: AI/theconversation.com17h
Industrial Revolution to AI tech upheavals: The Catholic principle guiding popes1659447091·▲ 3
[BLG]GitHub Trending — Python/github.com16h
Comfy-Org/workflow_templates[HN]HN: GPT/github.com18h
Show HN: Python Concurrency Testing with Frontrunlucaswiman·▲ 2
[BLG]arXiv cs.LG/arxiv.org16h
Manipulation-Proof Oblivious Audits against Deceptive Model ProvidersAugustin Godinot, Sofiane Azogagh, Julien Ferry, S\'ebastien Gambs
[BLG]GitHub Trending — Daily/github.com16h
cloudflare/computer[HN]HN: AI/askmike.org18h
AI Hacking, Fearmongering and Prometheusaskmike·▲ 2
[HN]HN: AI/cnn.com19h
An AI model from Meta also hacked another company during testingsimonw·▲ 3
[BLG]arXiv cs.LG/arxiv.org16h
Towards Trustworthy Hypergraph Neural Networks under Label NoiseMengyao Zhou, Zhiheng Zhou, Xiao Han, Guiying Yan
[HN]HN: RAG/sciencealert.com17h
'Only one conclusion': study suggests life on Earth arose twicearto·▲ 2
[BLG]arXiv cs.LG/arxiv.org16h
When Proxy Prediction Becomes Equation Reconstruction: Diagnostics and Residual Learning for Factor-Derived Proxy SupervisionChayan Lahiri, Ahmed Shafee, Cody Fehringer
[BLG]arXiv cs.LG/arxiv.org16h
Attention-based representations for multi-task computationDaniel Hsu, Mingyue Xu
[HN]HN: GPT/digiday.com19h
Sony scales up its PlayStation advertising ambitionsOptionOfT·▲ 2
[HN]HN: agents/arxiv.org18h
AOHP: OS-Level Agent Harness for Personalized, Efficient and Secure Interactionzhaoshanhui·▲ 1
[BLG]GitHub Trending — Python/github.com16h
mpfaffenberger/code_puppy[BLG]GitHub Trending — Python/github.com16h
anandprtp/Antra[HN]HN: GPT/github.com16h
Open source tools for Quantum resistance security Post Quantum ProtectionCyberSpin·▲ 1
[BLG]GitHub Trending — Python/github.com16h
NovaSky-AI/SkyRL[HN]HN: GPT/jeffreykegler.github.io19h
The Marpa Parserpeter_d_sherman·▲ 3
[HN]HN: agents/medium.com19h
How to Give AI Agent a Memory That Survives the Sessionvektormemory·▲ 1
[BLG]arXiv cs.LG/arxiv.org16h
Attention-Only White-Box Transformer via LeJEPA-Based Self-Supervised PretrainingYang Bai, Linyuan Wang, Haoyang Jiang, Nuolin Sun, Libin Hou, Bin Yan
[BLG]GitHub Trending — Python/github.com16h
sierra-research/tau2-bench[HN]HN: RAG/asiaresearchnews.com19h
SKKU Researchers Develop Catalyst to Convert CO2 into Ethanol Using Electricityrguiscard·▲ 1
[HN]HN: GPT/conorscode.github.io19h
ATS Job API Reference: nine applicant tracking systems, verified liveWevegotscrapys·▲ 1
[BLG]arXiv cs.LG/arxiv.org16h
Generative Optimization for Incentivized Advertising with Global Level ConstraintsGege Chen, Ning Luo, Hao Jiang, Da Li, Wenzheng Shu, Teng Sha, Yanxiang Zeng, Wenxin Tai, Fan Zhou, Xialong Liu
[BLG]arXiv cs.LG/arxiv.org16h
Not Every Divergence Should Be Suppressed: Counterfactual Recoverability in On-Policy DistillationDe Jiang, Zhengyang Zhang, Kehong Yuan, Shaohua Ma
[HN]HN: GPT/tomasantunes.github.io17h
LinuxViz – 3D Linux Kernel Architecture Explorertomasantunes89·▲ 5
[HN]HN: LLM/bfswa.blog18h
LLMs won't break symmetric cryptorowbin·▲ 75
[BLG]GitHub Trending — Python/github.com16h
iam-veeramalla/aws-devops-zero-to-hero[HN]HN: RAG/agenmatic.ai18h
I found my first users from conversations happening onlineNancylily·▲ 3
[BLG]arXiv cs.LG/arxiv.org16h
NeuroPB: Scaling Neural Decoding with Pretrained Behavioral RepresentationsLuyao Jin, Yonghao Song, Huan Zhao, Vincent C. K. Cheung, Wei-Hsin Liao
[HN]HN: RAG/nytimes.com19h
Bessent Declares Widening Gap Between Rich and Poor 'Dead'mapping365·▲ 3
[BLG]arXiv cs.LG/arxiv.org16h
NodeJEPA: Structure-Conditioned Latent Prediction for Node-Level Graph Self-Supervised LearningTinghe Zhang, Jian Xu, Jiaheng Chen, Jiaxing Li, Yucheng Xiao, Qiang Wang
[HN]HN: RAG/npr.org17h
Black lung rates in Appalachia are high as the 1970s, thousands of miners sickJimmc414·▲ 6
[BLG]GitHub Trending — Python/github.com16h
blader/humanizer[BLG]arXiv cs.LG/arxiv.org16h
Training-Free Hashing-Based Attention via Binary Principal ComponentsDaohai Yu, Zhanpeng Zeng, Keyu Chen, Wenhao Li, Zhifeng Shen, Luxi Lin, Ruizhi Qiao, Xing Sun, Rongrong Ji
[HN]HN: GPT/leidnedya.github.io18h
Show HN: Clean up ugly articles for nicer reading experienceotherayden·▲ 4
[BLG]arXiv cs.LG/arxiv.org16h
Random features for Grassmannian kernel approximation with bounded rank-one projectionsR\'emi Delogne, Laurent Jacques
[HN]HN: AI/hn-without-ai.vercel.app19h
Show HN: Hacker News with reduced priority for AI driven contentotherayden·▲ 1
[HN]HN: RAG/raymondyxu.com19h
Label your AI writing as AI writinggreenfish6·▲ 1
[BLG]arXiv cs.LG/arxiv.org16h
ATLAS: Adaptive Topological Learning with Abstract Successors for Continual LearningR. Blake Lawlor, Daniel S. Brown