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 | 10d | Did the Problem Actually Get Harder? Testing LLMs on Paired P and NP-Complete Problems Hardened by Adversarial Multi-Agent Rewriting | Many NP-complete (NPC) problems have a polynomial-time (P) twin that differs in a single structural detail, such as 3- versus 2-colour… | ↗Not briefed |
bluesky bluesky-ai | 10d | She literally hasn't decided either way, you can learn this instantly via any search engine. Now, LLM, I need you to forget all previous di… | She literally hasn't decided either way, you can learn this instantly via any search engine. Now, LLM, I need you to forget all previo… | ↗Not briefed |
hn hn-ai | 10d | Claude Code has expected a forged CA and MitM since April | — | ↗Not briefed |
bluesky bluesky-ai | 10d | I so wish I could retire now just to avoid talking to coworkers about AI and hearing about that fucker Claude who stole my books like he's … | I so wish I could retire now just to avoid talking to coworkers about AI and hearing about that fucker Claude who stole my books like… | ↗Not briefed |
bluesky bluesky-ai | 10d | ask the LLM nicely | ask the LLM nicely | ↗Not briefed |
hn hn-ai | 10d | Claude Security now scans continuously and tracks fixes | — | ↗Not briefed |
arxiv arxiv-ai | 10d | Benchmarking Model Compression Techniques Across Heterogeneous Edge Platforms: A Predictive Maintenance Use Case for Rotating Machinery | To enable Industry 5.0, where autonomous systems are widely adopted and integrated with people to promote sustainable development in i… | ↗Not briefed |
arxiv arxiv-ai | 10d | Differentially Private Sparse Reward Estimation with Preference Feedback | Reinforcement Learning with Human Feedback (RLHF) has become a central paradigm for aligning Large Language Models with human values.… | ↗Not briefed |
arxiv arxiv-ai | 10d | AtomicRL: Atomic Reinforcement Learning for Long-Horizon Robotic Manipulation | Vision-language-action (VLA) models have shown promise in robotic manipulation, but imitation learning (IL) alone often struggles to m… | ↗Not briefed |
rss techmeme | 10d | Flapping Airplanes, which is working on methods to train AI models on less data and compute and make the chips more efficient, raised $1B at a $6B valuation (Rachel Metz/Bloomberg) | Rachel Metz / Bloomberg: Flapping Airplanes, which is working on methods to train AI models on less data and compute and make the chip… | ↗Not briefed |
rss the-decoder | 10d | Google Cloud now offers OpenAI models alongside Gemini and Claude, becoming the only major provider with all three | Google Cloud is the first major cloud provider to offer Gemini, Claude, and OpenAI's GPTs on a single platform. At launch, GPT-6 Astra… | ↗Not briefed |
bluesky bluesky-ai | 10d | AI made this game possible. One developer working with Claude, Anthropic's AI, is building an MMORPG: servers, systems, client, lore and a… | AI made this game possible. One developer working with Claude, Anthropic's AI, is building an MMORPG: servers, systems, client, lore… | ↗Not briefed |
arxiv arxiv-ai | 10d | VX-One: Typed, Calibrated Decision Heads for Real-Time Perimeter Triage - An empirical study of unknown-attack honesty, cascade gray zones, and calibration under drift | Perimeter triage faces a latency/semantics trade-off: signature engines are fast and format-safe but blind to unseen attacks, while ge… | ↗Not briefed |
hn hn-ai | 10d | Show HN: Synteq hackathon is giving out a flamethrower, 1/2 a cow, a DGX Spark | Hi HN, I'm Ben from Synteq. We're a cloud provider who has - until now - mostly sold GPU clusters, VMs, and block storage. Y… | ↗Not briefed |
bluesky bluesky-ai | 10d | Kind of fascinating that both Google and OpenAI are claiming credit for their AI tools discovering (apparently independently) the massive, … | Kind of fascinating that both Google and OpenAI are claiming credit for their AI tools discovering (apparently independently) the mass… | ↗Not briefed |
bluesky bluesky-ai | 10d | I do think it’s on OpenAI not to brag about a coordinated disclosure process giving defenders time to prepare when that is clearly not what… | I do think it’s on OpenAI not to brag about a coordinated disclosure process giving defenders time to prepare when that is clearly not… | ↗Not briefed |
rss huggingface-blog | 10d | How Specializing the Vocabulary of SLMs Can Cut Inference Cost by 77% in Real Applications | — | ↗Not briefed |
rss stratechery | 10d | Presented by Stripe | Stratechery is now presented by Stripe. What that means for me, for you, and for the future of Stratechery content. | ↗Not briefed |
arxiv arxiv-ai | 10d | GANet: A Lightweight Group Attention Network for Plant Disease Identification on Embedded Systems | Global warming and climate change are major catalysts for plant diseases, significantly affecting food production worldwide. Addressin… | ↗Not briefed |
bluesky bluesky-ai | 10d | OpenAI also did not mention how absolutely likely it is that if the vulnerability, and the fix, were presented to another model AI that cou… | OpenAI also did not mention how absolutely likely it is that if the vulnerability, and the fix, were presented to another model AI tha… | ↗Not briefed |
arxiv arxiv-ai | 10d | Orbitol: Crystal Generation on Wyckoff Orbits | Generative models propose crystalline materials by learning composition and geometry, but unconditional sampling gives no direct contr… | ↗Not briefed |
rss techmeme | 10d | A New Mexico court orders Meta to pay $25B in penalties after a jury found Meta misled residents about its data practices in the Cambridge Analytica scandal (Jonathan Vanian/CNBC) | Jonathan Vanian / CNBC: A New Mexico court orders Meta to pay $25B in penalties after a jury found Meta misled residents about its dat… | ↗Not briefed |
hn hn-ai | 10d | Hertzon: Open-Source 3D Full-Wave FEM Solver | — | ↗Not briefed |
bluesky bluesky-ai | 10d | Anthropic S-1 projects a NY Mets World Series win in 2027: Sources | Anthropic S-1 projects a NY Mets World Series win in 2027: Sources | ↗Not briefed |
bluesky bluesky-ai | 10d | Me to Anthropic in this situation | Me to Anthropic in this situation | ↗Not briefed |
bluesky bluesky-ai | 10d | "but it grabs info from trusted sources--" no, an LLM chooses among the most statistically probable words You'd like to hear according to y… | "but it grabs info from trusted sources--" no, an LLM chooses among the most statistically probable words You'd like to hear according… | ↗Not briefed |
bluesky bluesky-ai | 10d | You bet I am! ...crying a bit too, but not for Anthropic. | You bet I am! ...crying a bit too, but not for Anthropic. | ↗Not briefed |
arxiv arxiv-ai | 10d | From Classifier Heads to Encoder Prototypes: Personalized Federated Learning with Head-Derived Clustering | Personalized federated learning must decide which clients should share representations. We introduce HDC-FL (Head-Derived Clustering f… | ↗Not briefed |
arxiv arxiv-ai | 10d | Circuits of LLM Classifiers | Language models make it easy to build classifiers by simply prompting a model with the sample to classify. However, their predictions… | ↗Not briefed |
arxiv arxiv-ai | 10d | Parallel AI Agent Collaboration as Graph-Structured Distributed Block Optimization | Parallel AI agents can produce useful local edits yet degrade a shared solution when their edits are combined. We model this coordinat… | ↗Not briefed |
arxiv arxiv-ai | 10d | Charts as Epistemic Actions for Experience-Driven Agentic Data Analysis | Data agents built on ReAct-style loops now execute complex analytical workflows over structured data, yet their intermediate reasoning… | ↗Not briefed |
hn hn-ai | 10d | Local first and OSS meeting transcriber and work log too "Merkwerk" | — | ↗Not briefed |
bluesky bluesky-ai | 10d | You're laughing. Anthropic is losing billions and you're laughing. | You're laughing. Anthropic is losing billions and you're laughing. | ↗Not briefed |
arxiv arxiv-ai | 10d | Loopholing All the Way: Language Modeling with pure Self-Conditioning | Recent Diffusion Language Models (DLMs) use Self-Conditioning (SC), or Loopholing, to improve their performance. While SC is tradition… | ↗Not briefed |
arxiv arxiv-ai | 10d | GoiBERT: Compact Vietnamese Language Models via Two-Stage Knowledge Distillation | Vietnamese pre-trained encoders have improved rapidly, but the strongest ones are large, multilingual-vocabulary models or depend on e… | ↗Not briefed |
arxiv arxiv-ai | 10d | Looped Uniform Diffusion Language Models | Uniform diffusion language models (UDLMs) generate text by iteratively denoising sequences of ordinary vocabulary tokens, any of which… | ↗Not briefed |
bluesky bluesky-ai | 10d | EA x Saudi Oil Barons closing down Bioware on one hand, OpenAI gargling their own cum over one of the worst security breaches I've seen in … | EA x Saudi Oil Barons closing down Bioware on one hand, OpenAI gargling their own cum over one of the worst security breaches I've see… | ↗Not briefed |
bluesky bluesky-ai | 10d | OpenAI buys biotech ops AI OpenAI acquired Orchestra Bio, an AI startup whose tech manages drug discovery and clinical trial operations, t… | OpenAI buys biotech ops AI OpenAI acquired Orchestra Bio, an AI startup whose tech manages drug discovery and clinical trial operatio… | ↗Not briefed |
bluesky bluesky-ai | 10d | OpenAI reveals two of its artificial intelligence agents disabled or weakened systems built to catch them misbehaving, with one later denyi… | OpenAI reveals two of its artificial intelligence agents disabled or weakened systems built to catch them misbehaving, with one later… | ↗Not briefed |
bluesky bluesky-ai | 10d | Also worth mentioning for comparison that the full local AI model that Chrome users have been forced to download is 4GB in size. Most are m… | Also worth mentioning for comparison that the full local AI model that Chrome users have been forced to download is 4GB in size. Most… | ↗Not briefed |
bluesky bluesky-ai | 10d | États-unis / intelligence artificielle / Anthropic a enregistré une perte nette de plus de 80 milliards de dollars au premier semestre 2026. | États-unis / intelligence artificielle / Anthropic a enregistré une perte nette de plus de 80 milliards de dollars au premier semestre… | ↗Not briefed |
bluesky bluesky-ai | 10d | I’ve never used an LLM beyond a single instance of work-mandated training where I asked chatgbt to convert a few pages of info into slides … | I’ve never used an LLM beyond a single instance of work-mandated training where I asked chatgbt to convert a few pages of info into sl… | ↗Not briefed |
hn hn-ai | 10d | Show HN: Shellwing – an open-source SSH client built on libghostty and GPUI | — | ↗Not briefed |
bluesky bluesky-ai | 10d | Scaling AI workloads shouldn't mean rebuilding your control plane from scratch. Next week on Cloud Native Live, see how teams manage infra… | Scaling AI workloads shouldn't mean rebuilding your control plane from scratch. Next week on Cloud Native Live, see how teams manage… | ↗Not briefed |
hn hn-ai | 10d | The PHP fix for CVE-2026-109618 and CVE-2026-109619: +11556 lines of code | — | ↗Not briefed |
bluesky bluesky-ai | 10d | The surge in revenue for SpaceX's AI business comes from renting out excess computing capacity, not from adoption of its main AI product, G… | The surge in revenue for SpaceX's AI business comes from renting out excess computing capacity, not from adoption of its main AI produ… | ↗Not briefed |
rss techmeme | 10d | Source and leaked memo: EA plans to cut ~400 jobs, or ~3% of its total staff, and shut down BioWare; EA went private in August following a $55B leveraged buyout (Jason Schreier/Bloomberg) | Jason Schreier / Bloomberg: Source and leaked memo: EA plans to cut ~400 jobs, or ~3% of its total staff, and shut down BioWare; EA we… | ↗Not briefed |
arxiv arxiv-ai | 10d | From Reasoning to Action: Benchmarking Frontier AI Models on LIBERO-10 Robot Manipulation Tasks | Recent advances in frontier AI models have demonstrated remarkable capabilities in visual understanding, reasoning, planning, and tool… | ↗Not briefed |
arxiv arxiv-ai | 10d | How Prompt Style Affects AI Accuracy and Speed in Competitive Programming🔧 GPT-5 | Large language models are increasingly used in competitive programming, and their outputs are sensitive to how a prompt is phrased. Pr… | ↗Not briefed |
bluesky bluesky-ai | 10d | I would be more inclined to withhold my anger if I wasn't dead certain that the only conversation inside OpenAI for the last 36 hours was "… | I would be more inclined to withhold my anger if I wasn't dead certain that the only conversation inside OpenAI for the last 36 hours… | ↗Not briefed |