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 | you know the LLM torture box? yeah, that, but for humans. | you know the LLM torture box? yeah, that, but for humans. | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Social SDK, a typed TypeScript toolkit for 7 social platforms | — | ↗Not briefed |
hn hn-ai | 1mo | Nexusyn – Long-term memory engine for AI agents (Go, DiskANN, MCP) | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Low-Cost Video--Time Priors as a Strong Baseline for EEG--fNIRS Emotion Regression on Familiar Videos | Continuous emotion regression estimates moment-to-moment valence and arousal while a viewer watches a video. In familiar-video deploym… | ↗Not briefed |
bluesky bluesky-ai | 1mo | nice chart of absolute levels vs share too. if this is representative, Anthropic's ARR is declining. this would be a huge deal but seems to… | nice chart of absolute levels vs share too. if this is representative, Anthropic's ARR is declining. this would be a huge deal but see… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Offenbar kam es im Vorfeld der Präsentation der Enzyklika „Magnifica humanitas“ hinter den Kulissen zwischen Papst Leo XIV. und dem Technol… | Offenbar kam es im Vorfeld der Präsentation der Enzyklika „Magnifica humanitas“ hinter den Kulissen zwischen Papst Leo XIV. und dem Te… | ↗Not briefed |
hn hn-ai | 1mo | Anthropic invests $100M to train 10k engineers | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | I'd suggest that the reason the LLM companies use relatively rare or infrequent events as examples is that when people are well-versed in a… | I'd suggest that the reason the LLM companies use relatively rare or infrequent events as examples is that when people are well-versed… | ↗Not briefed |
hn hn-ai | 1mo | Using light-dark() without getting caught out | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Mastering Atari 2600 Games with Discovered Options | Temporal abstractions, often instantiated as options, have long been regarded as a mechanism for accelerating credit assignment, facil… | ↗Not briefed |
bluesky bluesky-ai | 1mo | lobbying theologians to agree with LLM consciousness is baffling to be so far up your own ass to try and get the catholics to agree with "… | lobbying theologians to agree with LLM consciousness is baffling to be so far up your own ass to try and get the catholics to agree w… | ↗Not briefed |
bluesky bluesky-ai | 1mo | ‘The aim is to make the AI assistants a little more approachable, distinct and trustworthy. If the Jolly Green Giant could get kids to eat … | ‘The aim is to make the AI assistants a little more approachable, distinct and trustworthy. If the Jolly Green Giant could get kids to… | ↗Not briefed |
arxiv arxiv-ai | 1mo | When a Correct Reward Is Not Enough: Diagnosing and Guiding PPO in an Analytically Solved Broker-Trader Game | Reinforcement learning (RL) is increasingly used for financial optimal-control problems when complex dynamics make analytical strategi… | ↗Not briefed |
bluesky bluesky-ai | 1mo | I recently had an email exchange in which someone explained they'd had an invigorating intellectual "argument" (their word) about some hist… | I recently had an email exchange in which someone explained they'd had an invigorating intellectual "argument" (their word) about some… | ↗Not briefed |
arxiv arxiv-ai | 1mo | A Path Integral Surrogate for Multi-Step Gradient Inversion in Federated Learning | Federated learning lets many clients train a shared model together without ever sending their private data to a central server. Each c… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Two Feet a tiny skate simulator in the browser | I made a small browser-based skate simulator called Two Feet.<p>Would love feedback on the controls and physics<p><a href="https:/… | ↗Not briefed |
rss techcrunch-ai | 1mo | Circuit Breaker Labs hopes to make AI safer for your kids (and you) | With all the talk about how AI might one day ill us all, it's easy to forget that AI has already harmed some people, psychologically.… | ↗Not briefed |
rss stratechery | 1mo | 2026.40: Dots and Question Marks | The best Stratechery content from the week of September 28, 2026, including Meta's focus, what OpenAI is doing, and Mao and NBA Media… | ↗Not briefed |
bluesky bluesky-ai | 1mo | i suppose Anthropic *could* all just be one elaborate Andy Kaufman bit | i suppose Anthropic *could* all just be one elaborate Andy Kaufman bit | ↗Not briefed |
arxiv arxiv-ai | 1mo | HazardWeaver: Scientific Route Selection for Hazard Analysis Agents | Understanding and assessing natural hazards is essential for disaster preparedness and risk reduction. Recent advances in large langua… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Science Fiction and Fantasy Writers Association (SFWA) and ten other writers' groups sign on to Authors Guild's letter to publishers urging… | Science Fiction and Fantasy Writers Association (SFWA) and ten other writers' groups sign on to Authors Guild's letter to publishers u… | ↗Not briefed |
bluesky bluesky-ai | 1mo | I basically run a little Lovecraft thinking in the background when I'm talking to you LLM folks. | I basically run a little Lovecraft thinking in the background when I'm talking to you LLM folks. | ↗Not briefed |
bluesky bluesky-ai | 1mo | They want AI personhood so they have a scapegoat for all their crimes. Claude is just a liability protection. | They want AI personhood so they have a scapegoat for all their crimes. Claude is just a liability protection. | ↗Not briefed |
arxiv arxiv-ai | 1mo | Threat-Preserving Representation Sensitivity in Agent-Security Benchmarks🔧 GPT-5 | Security benchmarks for LLM-based agents often report the attack success rate (ASR) as a measure of model robustness and use these sco… | ↗Not briefed |
bluesky bluesky-ai | 1mo | So Anthropic is anthropomorphizing its product? Who’d have guessed? | So Anthropic is anthropomorphizing its product? Who’d have guessed? | ↗Not briefed |
bluesky bluesky-ai | 1mo | AI guys have a God complex so out of control they think the Pope himself will validate their beliefs. | AI guys have a God complex so out of control they think the Pope himself will validate their beliefs. | ↗Not briefed |
bluesky bluesky-ai | 1mo | OpenAI sacks safety researchers for sharing too much information with other safety researchers, after OpenAI has been displayed to be delib… | OpenAI sacks safety researchers for sharing too much information with other safety researchers, after OpenAI has been displayed to be… | ↗Not briefed |
bluesky bluesky-ai | 1mo | I took the claude “Interview” this morning and the stance of author of the questions is so clearly “AI is a moral actor” that I ended up ju… | I took the claude “Interview” this morning and the stance of author of the questions is so clearly “AI is a moral actor” that I ended… | ↗Not briefed |
bluesky bluesky-ai | 1mo | I think people talking about how they're clearly only in it for the money don't necessarily realize how rich they already were before start… | I think people talking about how they're clearly only in it for the money don't necessarily realize how rich they already were before… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Rethinking What to Cache in Few-Step Diffusion Transformers: Solver-Aware Target Selection🔧 FLUX | Diffusion Transformers (DiTs) can generate high-quality images and videos, but generating each sample requires multiple costly DiT for… | ↗Not briefed |
bluesky bluesky-ai | 1mo | according to the Other Place, David Robinson (head of policy planning and safety systems lead) left openai the day the safety folks got fir… | according to the Other Place, David Robinson (head of policy planning and safety systems lead) left openai the day the safety folks go… | ↗Not briefed |
hn hn-ai | 1mo | Minecraft is Getting a New Dimension [video] | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | HyperBrowseComp: A Multilingual and Multimodal Stress Test for Web-Browsing Agents | We introduce HyperBrowseComp, a multilingual and multimodal browsing benchmark comprising 423 manually authored and human-validated qu… | ↗Not briefed |
bluesky bluesky-ai | 1mo | "But have you used my latest favorite LLM?" and "I'm sure the problem was the input you fed into the LLM and not the LLM itself" (2/2) | "But have you used my latest favorite LLM?" and "I'm sure the problem was the input you fed into the LLM and not the LLM itself" (2/2) | ↗Not briefed |
hn hn-ai | 1mo | Trillium Labs (Nathan Lambert's new effort) | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | gen AI in videogame modding is the new milkshake duck for me. Holy shit somebody made Project Zomboid 3D! we regret to inform you two diffe… | gen AI in videogame modding is the new milkshake duck for me. Holy shit somebody made Project Zomboid 3D! we regret to inform you two… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Learning to Assess Heartbeat Observability for mmWave Heart-Rate Sensing | Contactless heart-rate sensing with millimeter-wave (mmWave) radar requires assessing whether individual measurements support reliable… | ↗Not briefed |
hn hn-ai | 1mo | SF's first public Mandarin immersion school gets a temporary campus | — | ↗Not briefed |
hn hn-ai | 1mo | OpenAI Deprecations | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Writerslogic at the CLEF 2026 SimpleText Track: Multi-Candidate LLM Simplification and Stacked Complexity Spotting | We describe the Writerslogic team's participation in the CLEF 2026 SimpleText shared task, addressing Task 1 (text simplification) and… | ↗Not briefed |
arxiv arxiv-ai | 1mo | Writerslogic at PAN 2026: Process over Content for Robust Detection under Domain Shift | We describe the Writerslogic systems for three PAN at CLEF 2026 shared tasks (Reasoning Trajectory Detection, Voight-Kampff Generative… | ↗Not briefed |
hn hn-ai | 1mo | OpenAI's Astra model is shockingly good at robotics | — | ↗Not briefed |
arxiv arxiv-ai | 1mo | Knowledge or Calculator? Decomposing the Skill Premium in Verifiable Financial Agent Workflows | Financial AI agents must do more than retrieve facts: investment workflows require correct quantitative execution, reliable use of pro… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Look, I'm not saying you CAN'T lobby The Pope, but you want to be I don't know King Philip II of Spain wanting your daughter's volatile and… | Look, I'm not saying you CAN'T lobby The Pope, but you want to be I don't know King Philip II of Spain wanting your daughter's volatil… | ↗Not briefed |
bluesky bluesky-ai | 1mo | I'm not entirely sure why Anthropic wants this. To say they aren't liable for anything the other being did? But even regular tort law doesn… | I'm not entirely sure why Anthropic wants this. To say they aren't liable for anything the other being did? But even regular tort law… | ↗Not briefed |
bluesky bluesky-ai | 1mo | It’s also the broad categorization of *anything* we wouldn’t have called AI ten years ago as AI nowadays. Matching learning and visual patt… | It’s also the broad categorization of *anything* we wouldn’t have called AI ten years ago as AI nowadays. Matching learning and visual… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: GPT-6 Astra and Claude 5.5 Opus race to create the best StarCraft bot | StarSkirmish: Hillclimb is a long horizon implementation of the StarCraft: Broodwar based LLM benchmark that I released last week.<p>E… | ↗Not briefed |
bluesky bluesky-ai | 1mo | Although generative AI systems rely heavily on Wikipedia's vast trove of information, some companies won't pay for it, the CEO said. Publ… | Although generative AI systems rely heavily on Wikipedia's vast trove of information, some companies won't pay for it, the CEO said.… | ↗Not briefed |
hn hn-ai | 1mo | Show HN: Odd Essentials Agent Guild | — | ↗Not briefed |
bluesky bluesky-ai | 1mo | A new PS1 recomp comes out with a disclaimer, and you brace yourself because you feel this is about to be another "we used AI", but it's ac… | A new PS1 recomp comes out with a disclaimer, and you brace yourself because you feel this is about to be another "we used AI", but it… | ↗Not briefed |