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 | Guide, Then Let Go: Gap-Adaptive Teacher Scheduling for Sparse-Reward Agentic RL | Reinforcement learning for long-horizon agents typically relies on sparse outcome-based rewards. This leads to a severe cold-start pro… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Everyone clowning on Anthropic for noting its product is 'an existential threat to humanity' in ITS OWN IPO filing. But it doesn't matter.… | Everyone clowning on Anthropic for noting its product is 'an existential threat to humanity' in ITS OWN IPO filing. But it doesn't ma… | ↗Not briefed |
rss theverge-ai | 1mo | OpenAI DevDay 2026: The biggest news and announcements🔧 Hugging Face | It’s OpenAI’s turn in the fall tech events calendar. The company is hosting its annual DevDay on September 29th in San Francisco, star… | ↗Not briefed |
arxiv arxiv-ai | 1mo | It's All Training: A Fully Synthetic Single-Stage Recipe for LLMs | Current pre-training datasets are derived from web crawls, with all their issues, and were not designed to support mid- and post-train… | ↗Not briefed |
arxiv arxiv-ai | 1mo | ReCAP: Retrieval-Guided Capability Reuse for Multimodal Continual Instruction Tuning | Multimodal continual instruction tuning (MCIT) aims to enable multimodal large language models to acquire new capabilities from sequen… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Visual Branch is What You Need for CLIP-based Class-Incremental Learning | Class-Incremental Learning (CIL) requires models to recognize new classes over time without forgetting previously learned ones. With t… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Behavioral Capacity Certificates for Quantized Language Models | Activation and key-value cache precision change what a quantized language model computes without altering its stored weights. Direct w… | ↗Not briefed |
hn hn-ai | 1mo | EFF to SFPD: Drones Are Powerful Surveillance Tools That Require a Robust Policy | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Boids of a Feather Flock Together - Evolving Prey Behaviours Under Different Predator Attack Strategies | Flocking and schooling are thought to have evolved partly as defences against predation, but how prey should balance social and escape… | ↗Not briefed |
rss simon-willison | 1mo | OpenAI DevDay 2026 live blog | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | TopoEmbedX: A General Framework for Representation Learning on Topological Domains | Topological structures such as simplicial complexes, hypergraphs, and cell complexes extend standard graph models by modeling higher-o… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Zero-shot Dependency Parsing with Unsupervised Cross-Lingual Bootstrapping | Pre-trained language models (PLMs) with encoder-based architectures have shown impressive capabilities in zero-shot cross-lingual tran… | ↗Not briefed |
arxiv arxiv-ai | 1mo | How Many Labels Does a Language Need? Annotation Budgets and Cross-Lingual Pooling for African-Language Text Classification | Every text classifier for an African language begins with a budgeting question: how many labelled examples are needed, and can labels… | ↗Not briefed |
bluesky bluesky-ai | 1mo | AI is burning through money, regs, energy, water - all for some foggy promise of what frankly sounds like a dystopian future. Reuters: | AI is burning through money, regs, energy, water - all for some foggy promise of what frankly sounds like a dystopian future. Reuters: | ↗Not briefed |
arxiv arxiv-ai | 1mo | Retrieval Capacity of Self-Attention Under Competition | How many tokens from its context does a language model actually use, and what determines that number? We study this question through s… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Estudio publicado en Nature analizó 1,4 millones de imágenes y vídeos (de Google, Wikipedia, IMDb, Flickr y YouTube) + 9 LLM y encontró que… | Estudio publicado en Nature analizó 1,4 millones de imágenes y vídeos (de Google, Wikipedia, IMDb, Flickr y YouTube) + 9 LLM y encontr… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Co-PiLOT: Constrained Physics-Informed Latent Optimization for Target-Driven Inverse Design | Inverse design of physical systems (molecules, devices, microstructures) often reduces to optimizing a high-dimensional structure agai… | ↗Not briefed |
rss bing-news-ai | 1mo | Top AI executives sign commitment to ‘self-police’ after meeting at White House | President Donald Trump said Tuesday that top tech executives had signed a new commitment aimed at better “self-policing” their compani… | ↗Not briefed |
rss bing-news-ai | 1mo | Top AI executives sign commitment to ‘self-police’ after meeting at White House | President Donald Trump said Tuesday that top tech executives had signed a new commitment aimed at better “self-policing” their compani… | ↗Not briefed |
bluesky bluesky-ai | 1mo | ‘Tech giants like Amazon and Google hold huge stakes in Anthropic, which sit on their balance sheets at high valuations. If Anthropic delay… | ‘Tech giants like Amazon and Google hold huge stakes in Anthropic, which sit on their balance sheets at high valuations. If Anthropic… | ↗Not briefed |
arxiv arxiv-ai | 1mo | ExceptionDrive: A Planning-Oriented Counterfactual Corner-Case Benchmark for Autonomous Driving | Average performance on routine driving benchmarks does not establish planner reliability under rare, safety-critical hazards. We propo… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Learning Beyond What You Sample: Off-Policy-Aware Cross-Model Trajectory Exchange for RLVR | Reinforcement Learning with Verifiable Rewards (RLVR) methods such as GRPO rely on successful self-generated trajectories, but finite… | ↗Not briefed |
hn hn-ai | 1mo | Bill Gates's Blunt Warning on A.I | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | * An LLM is not a cogniser and the agent as a whole system is not a cogniser either. However an agent can be a decider, in the sense that t… | * An LLM is not a cogniser and the agent as a whole system is not a cogniser either. However an agent can be a decider, in the sense t… | ↗Not briefed |
bluesky bluesky-ai | 1mo | What is the practical outcome of this philosophising that Will thinks is irrelevant? OpenAI, Anthropic, et al want you to believe that the… | What is the practical outcome of this philosophising that Will thinks is irrelevant? OpenAI, Anthropic, et al want you to believe tha… | ↗Not briefed |
bluesky bluesky-ai | 1mo | But OpenAI, AI boosters, and the industry generally, want you to look away from the ontology, and have largely forgotten it themselves. The… | But OpenAI, AI boosters, and the industry generally, want you to look away from the ontology, and have largely forgotten it themselves… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Strict-Saddle Landscapes and Multi-Rank Geometry in Low-Tubal-Rank Tensor Sensing | We study the optimization landscape of low-tubal-rank tensor sensing through a balanced factorization. Under a tubal restricted isomet… | ↗Not briefed |
arxiv arxiv-ai | 1mo | AgentBug-Smith: Automatically Reproducing Real-World Harness Bugs in Agentic Systems | Agent harness bugs exhibit unique characteristics and remain challenging for state-of-the-art software agents to repair. Progress in t… | ↗Not briefed |
arxiv arxiv-ai | 1mo | It's Not What the Image Shows: Irrelevant Context Destabilises VLM Judges Without Informing Them | Vision-language models (VLMs) are increasingly used in place of human annotators, making it important that substitutability tests refl… | ↗Not briefed |
arxiv arxiv-ai | 1mo | One Threshold Does Not Fit All Languages: Language-Conditional Deferral for Reliable and Efficient Low-Resource Text Classification | In the Global South, the lower-income countries of Africa, Asia, and Latin America where most of the world's languages are spoken, a d… | ↗Not briefed |
bluesky bluesky-ai | 1mo | With one deft impression, the “Saturday Night Live” season premiere captured the dissonance between how tech CEOs and the public talk about… | With one deft impression, the “Saturday Night Live” season premiere captured the dissonance between how tech CEOs and the public talk… | ↗Not briefed |
rss techmeme | 1mo | IPO prospectus: Anthropic has agreements to pay SpaceX up to $84.5B to use its Nvidia-based compute through 2029, mostly cancellable with a 90-day notice period (Grace Kay/The Information) | Grace Kay / The Information: IPO prospectus: Anthropic has agreements to pay SpaceX up to $84.5B to use its Nvidia-based compute throu… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Storage Is Not Strategy: State-Conditioned Support Control for LLM Unlearning | Many localized large language model (LLM) unlearning methods select a small parameter subset from a localization signal and keep it fi… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Active Budget Can Kill Sensitivity: Diagnosing and Repairing TopK Sparse Autoencoder Reliability | Sparse autoencoders (SAEs) are increasingly scaled to wider dictionaries to recover fine-grained structure from large language model a… | ↗Not briefed |
hn hn-ai | 1mo | New PlayStation 5 Console Jailbreak Released | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | HandAnthro: Automated Hand Anthropometry from a Single Image | Hand anthropometry supports protective-glove design, but existing measurement methods often require trained operators, specialized har… | ↗Not briefed |
hn hn-ai | 1mo | Spirula-studio: 3D Gaussian Splatting trainer video to splat to mesh | — | ↗Not briefed |
hn hn-ai | 1mo | AI will transform capitalism – but how? | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | AnthroDial: Benchmarking LLM Anthropomorphism in Autonomous Social Interaction | Large language models (LLMs) are increasingly deployed as social agents, yet credible human-like interaction requires more than fluent… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Delta-Matching: Closing the Final Gap of Native 8-bit Training for LLMs | Reliable FP8 attention remains a barrier to fully native 8-bit large language model training. We derive how forward-backward inconsist… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: A terminal UI for GitHub Actions across all repos and orgs | We have many repos across multiple GitHub organizations (no monorepo yet). I needed a tool to list and manage Actions from the termina… | ↗Not briefed |
arxiv arxiv-ai | 1mo | FlowMap-OPD: Rollout--Kernel Separation for On-Policy Distillation of Few-Step Flow-Map Generators | Few-step flow-map generators, including MeanFlow and consistency models, enable efficient sampling through long-range transport, yet t… | ↗Not briefed |
bluesky bluesky-ai | 1mo | There's a lot of information (open-source and otherwise) about the difference between an LLM and a classifier in terms of what they are, ho… | There's a lot of information (open-source and otherwise) about the difference between an LLM and a classifier in terms of what they ar… | ↗Not briefed |
bluesky bluesky-ai | 1mo | The Manhattan Project cost $36B in today’s dollars. Two+ years to go from an empty ranch to nuclear fission, building on decades of Nobels …🔧 Consensus | The Manhattan Project cost $36B in today’s dollars. Two+ years to go from an empty ranch to nuclear fission, building on decades of No… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Is manual software optimization a thing of the past? | Scientific software is increasingly required to process larger datasets while maintaining acceptable execution times. Software optimiz… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Evaluation Choices Shape Biomedical ML Claims: A Pediatric Pneumonia Benchmark Case Study | Biomedical machine learning papers often compress model performance into one headline number. That number can look like a property of… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Scaling Influence Functions in LLMs through Eigenbasis-Corrected One-Bit Gradient Projection | Influence functions estimate how individual training examples affect the behavior of large language models (LLMs). Analyzing how train… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Counterfactual Probing for Parallel Unmasking with Hidden Forest Structure | Masked generative models offer parallel token prediction, but accurate parallel sampling must account for dependencies among tokens. W… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Estela – Rebuild billable hours from your Claude Code and Codex logs | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | because OpenAI is absolutely Pro-Rape | because OpenAI is absolutely Pro-Rape | ↗Not briefed |