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 | 17d | Anthropic logged net loss of over $80 billion in first half of 2026, WSJ reports reut.rs/4z2Zi9J | Anthropic logged net loss of over $80 billion in first half of 2026, WSJ reports reut.rs/4z2Zi9J | ↗Not briefed |
bluesky bluesky-ai | 17d | Here I was thinking this was some incredible AI exploit driven heist, that it was a real sign things were heating up and we need to work ev… | Here I was thinking this was some incredible AI exploit driven heist, that it was a real sign things were heating up and we need to wo… | ↗Not briefed |
hn hn-ai | 17d | Ask HN: Will AI = extreme concentration of power? | Let's say that AI will continue to get better without being a major risk due to loss of control. I feel like in this paradigm, it… | ↗Not briefed |
bluesky bluesky-ai | 17d | 2nd live of the day. LIVE NATIONWIDE AI WALKOUT | LIVE From OpenAI in NYC. Tune in 5:15pm ET www.youtube.com/live/18re9zX... | 2nd live of the day. LIVE NATIONWIDE AI WALKOUT | LIVE From OpenAI in NYC. Tune in 5:15pm ET www.youtube.com/live/18re9zX... | ↗Not briefed |
arxiv arxiv-ai | 17d | Minimax-Optimal Predictive Flow Matching under Distribution Drift | Flow matching learns a generative velocity field by regression, but under distribution drift a field must be deployed before the curre… | ↗Not briefed |
bluesky bluesky-ai | 17d | Reminder that the $80bn Anthropic net loss figure is going to be 95% driven by warrant liability accounting, an economically meaningless no… | Reminder that the $80bn Anthropic net loss figure is going to be 95% driven by warrant liability accounting, an economically meaningle… | ↗Not briefed |
hn hn-ai | 17d | Show HN: Agent-Zero – Starter template for AI-native coding projects | Hi All,<p>Agent-Zero is a GitHub template that you can use for new coding projects that use AI-first development workflows. It include… | ↗Not briefed |
arxiv arxiv-ai | 17d | Is Refusal Direction Universal across Low-Resourced Languages?🔧 Replicate | Refusal directions have been shown to transfer cross-lingually with near-perfect effectiveness across safety-aligned languages (Wang e… | ↗Not briefed |
arxiv arxiv-ai | 17d | CSD: Common-Support Distillation for Cross-Tokenizer Knowledge Transfer | Cross-tokenizer logits distillation enables knowledge transfer across model families, allowing students to learn from stronger or more… | ↗Not briefed |
bluesky bluesky-ai | 17d | The pilot agent actually writes code to manipulate the goals and constraints of lower-level control systems, which I think is pretty cool: … | The pilot agent actually writes code to manipulate the goals and constraints of lower-level control systems, which I think is pretty c… | ↗Not briefed |
bluesky bluesky-ai | 17d | I think it's probably mostly from Anthropic | I think it's probably mostly from Anthropic | ↗Not briefed |
bluesky bluesky-ai | 17d | Anthropic IPO | Anthropic IPO | ↗Not briefed |
hn hn-ai | 17d | Emudrahahahahaha | — | ↗Not briefed |
arxiv arxiv-ai | 17d | Combating Catastrophic Forgetting in Continual Domain Adaptation via Knowledge Recasting | Adapting language models to new domains during post-training risks degrading previously acquired capabilities, a phenomenon commonly k… | ↗Not briefed |
bluesky bluesky-ai | 17d | If people think a chatbot has a political bias, it becomes less effective AI chatbots effectively corrected misconceptions, but a short m… | If people think a chatbot has a political bias, it becomes less effective AI chatbots effectively corrected misconceptions, but a sh… | ↗Not briefed |
arxiv arxiv-ai | 17d | Fig: Automated Multiagent Behavior Elicitor🔧 Elicit | Many social evaluations for multiagent systems study behaviors as they emerge from environments instead of systematically intervening… | ↗Not briefed |
rss techmeme | 17d | Netflix says despite investor concern about its view-hours growth, its engagement is "healthy" and it will try to bring more third-party content to its service (Nicole Sperling/New York Times) | Nicole Sperling / New York Times: Netflix says despite investor concern about its view-hours growth, its engagement is “healthy” and i… | ↗Not briefed |
bluesky bluesky-ai | 17d | Earlier this year, Amodei went on Dwarkesh's podcast and mentioned how Anthropic had been growing ARRs tenfold every year for three years, … | Earlier this year, Amodei went on Dwarkesh's podcast and mentioned how Anthropic had been growing ARRs tenfold every year for three ye… | ↗Not briefed |
bluesky bluesky-ai | 17d | I liked the idea of this, but thought it could be far simpler AND still useful, so crafted a Claude in Chrom Skill + Bookmarklet version: a… | I liked the idea of this, but thought it could be far simpler AND still useful, so crafted a Claude in Chrom Skill + Bookmarklet versi… | ↗Not briefed |
hn hn-ai | 17d | Ask HN: Are agent benchmarks the new credit ratings? | I did some talking to engineers at new-ish startups and agent companies are buying benchmarks to prove reliability to their customers.… | ↗Not briefed |
rss techmeme | 17d | Netflix reports Q3 revenue up 12% YoY to $12.87B, vs. $12.87B est., net income up 44% to $3.66B, due to the timing of expenses; NFLX drops 4%+ after hours (Sarah Whitten/CNBC) | Sarah Whitten / CNBC: Netflix reports Q3 revenue up 12% YoY to $12.87B, vs. $12.87B est., net income up 44% to $3.66B, due to the timi… | ↗Not briefed |
hn hn-ai | 17d | Anthropic logged net loss of over $80B in first half of 2026 | — | ↗Not briefed |
bluesky bluesky-ai | 17d | "Anthropic S-1" (WSJ) "~1/5 of $16.3bn in revenue in the H1 of 2026 from just 2 customers" Plans to spend: $111bn on $GOOG cloud computin… | "Anthropic S-1" (WSJ) "~1/5 of $16.3bn in revenue in the H1 of 2026 from just 2 customers" Plans to spend: $111bn on $GOOG cloud com… | ↗Not briefed |
arxiv arxiv-ai | 17d | Building AI Research Teams for Autonomous Discovery of Linear Attention Architectures | Autonomous agents can increasingly propose ideas, modify code, and run experiments, creating an opportunity to accelerate scientific d… | ↗Not briefed |
bluesky bluesky-ai | 17d | No skills used anywhere, just prompts telling it to build with Three.js - though it did have initial sprites that it had generated with Ope… | No skills used anywhere, just prompts telling it to build with Three.js - though it did have initial sprites that it had generated wit… | ↗Not briefed |
hn hn-ai | 17d | Symbolic Learning in Python and Julia | — | ↗Not briefed |
hn hn-ai | 17d | nchat – Terminal-based messaging client | — | ↗Not briefed |
bluesky bluesky-ai | 17d | basically if you gave away 10% of your house to fund the down-payment, and then the value of your house skyrocketed, then you incurred "a l… | basically if you gave away 10% of your house to fund the down-payment, and then the value of your house skyrocketed, then you incurred… | ↗Not briefed |
bluesky bluesky-ai | 17d | How many deaths is an AI model worth? Anthropic’s draft bill would let the federal government block a model’s release only if it risks 5… | How many deaths is an AI model worth? Anthropic’s draft bill would let the federal government block a model’s release only if it ri… | ↗Not briefed |
arxiv arxiv-ai | 17d | SANE: Automated Scorer-Specific Metamorphic Testing for LLM Judges | LLM judges increasingly mediate benchmark evaluation, model selection, data curation, and training. Their decisions are useful only wh… | ↗Not briefed |
arxiv arxiv-ai | 17d | RL-IF: Tracking Non-local Training Data Influence for Online Reinforcement Learning | Online reinforcement learning (RL) has achieved significant success in decision-making tasks, but modern online RL remains highly sens… | ↗Not briefed |
bluesky bluesky-ai | 17d | *ANTHROPIC COULD SPEND $750B ON OBTAINING COMPUTING CAPACITY:WSJ *ANTHROPIC BOOKS $80.7B NET LOSS IN 1H ON ACCOUNTING CHARGES:WSJ jesus | *ANTHROPIC COULD SPEND $750B ON OBTAINING COMPUTING CAPACITY:WSJ *ANTHROPIC BOOKS $80.7B NET LOSS IN 1H ON ACCOUNTING CHARGES:WSJ je… | ↗Not briefed |
arxiv arxiv-ai | 17d | inr2action: Encoding Robot Episodes as Neural Implicit Representations for Recall and Action | Long-horizon robotic operation requires memory that scales with essential information rather than execution time. We present a paramet… | ↗Not briefed |
bluesky bluesky-ai | 17d | the $80B is a number that mechanically goes up with their valuation. cf. reuters on their $42B net loss in 2025 www.reuters.com/business/fi… | the $80B is a number that mechanically goes up with their valuation. cf. reuters on their $42B net loss in 2025 www.reuters.com/busine… | ↗Not briefed |
bluesky bluesky-ai | 17d | The entire "AI" industry is running on debt and venture capital. The venture capital is getting leery, neither Anthropic or OpenAI feel bra… | The entire "AI" industry is running on debt and venture capital. The venture capital is getting leery, neither Anthropic or OpenAI fee… | ↗Not briefed |
bluesky bluesky-ai | 17d | good thread years ago it was understood and openly admitted to be simply reselling other people's life work without compensating anyone fo… | good thread years ago it was understood and openly admitted to be simply reselling other people's life work without compensating anyo… | ↗Not briefed |
bluesky bluesky-ai | 17d | ANTHROPIC: Sets $80 billion on fire EXUBERANT INVESTORS: "Sign me up!" | ANTHROPIC: Sets $80 billion on fire EXUBERANT INVESTORS: "Sign me up!" | ↗Not briefed |
arxiv arxiv-ai | 17d | Actually, All Learnable Distribution Classes Are Privately Learnable | We show that every class of distributions that is learnable up to arbitrarily small target error in total variation distance with fini… | ↗Not briefed |
bluesky bluesky-ai | 17d | A neat natural experiment appears. DeepSeek is rate limited and gets skipped. Argon, to my surprise, seamlessly steps in to play the DeepSe… | A neat natural experiment appears. DeepSeek is rate limited and gets skipped. Argon, to my surprise, seamlessly steps in to play the D… | ↗Not briefed |
bluesky bluesky-ai | 17d | ANTHROPIC POSTED A NET LOSS EXCEEDING $80 BLN IN THE FIRST HALF OF THIS YEAR - WSJ | ANTHROPIC POSTED A NET LOSS EXCEEDING $80 BLN IN THE FIRST HALF OF THIS YEAR - WSJ | ↗Not briefed |
bluesky bluesky-ai | 17d | ‘The company has signaled that it could potentially spend $750 billion in the coming years on obtaining more computing capacity, including … | ‘The company has signaled that it could potentially spend $750 billion in the coming years on obtaining more computing capacity, inclu… | ↗Not briefed |
hn hn-ai | 17d | Get Infomaniak calendar directly in Firefox | — | ↗Not briefed |
arxiv arxiv-ai | 17d | To Manipulate or Not: Opinion Dynamics in LLM Interactions | Large language models increasingly take part in extended conversations where they can both shape and be shaped by another party's view… | ↗Not briefed |
rss arstechnica-tech | 17d | The Internet goes on high alert after discovery of 8 compromised TLS certs | Intermediate certs allow anyone to mint web credentials for any Internet server. | ↗Not briefed |
rss techmeme | 17d | Anthropic's IPO prospectus: H1 2026 revenue of $16.3B, ~20% from two customers, vs. $1.1B in H1 2025, and an $80.7B net loss driven by accounting charges (Anissa Gardizy/Wall Street Journal) | Anissa Gardizy / Wall Street Journal: Anthropic's IPO prospectus: H1 2026 revenue of $16.3B, ~20% from two customers, vs. $1.1B in H1… | ↗Not briefed |
rss techmeme | 17d | A profile of Xbox CEO Asha Sharma; sources: Xbox plans a major Game Pass overhaul, a next-gen handheld device by late 2027, and next-gen console by late 2028 (Sebastian Herrera/Fortune) | Sebastian Herrera / Fortune: A profile of Xbox CEO Asha Sharma; sources: Xbox plans a major Game Pass overhaul, a next-gen handheld de… | ↗Not briefed |
arxiv arxiv-ai | 17d | LEAN: Loss-Engineered Modulation of the Model Weight-Sensitivity Distribution for Efficient, Error-Tolerant Neural Networks | The weights of a trained network do not have uniform sensitivity to model accuracy: the 1% most sensitive weights carry about a fifth… | ↗Not briefed |
bluesky bluesky-ai | 17d | if it's news to you in October 2026 that Anthropic is positively blasting out tokens to a rapidly growing pool of customers I genuinely don… | if it's news to you in October 2026 that Anthropic is positively blasting out tokens to a rapidly growing pool of customers I genuinel… | ↗Not briefed |
bluesky bluesky-ai | 17d | I built a Command & Conquer style RTS called Battle Drones, mainly using Claude Opus 5.5, which runs in the browser and includes device-to-… | I built a Command & Conquer style RTS called Battle Drones, mainly using Claude Opus 5.5, which runs in the browser and includes devic… | ↗Not briefed |
arxiv arxiv-ai | 17d | How Much Gradient Does Truncated Backprop Lose in Looped Transformers? | Looped transformers reuse one block of layers many times and are often trained by backpropagating through only the last k loops. We me… | ↗Not briefed |