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 | 2mo | Adtech should be regulated, frankly, out of existence to begin with. But using an LLM to generate or serve ads should give you jail time. I… | Adtech should be regulated, frankly, out of existence to begin with. But using an LLM to generate or serve ads should give you jail ti… | ↗Not briefed |
arxiv arxiv-ai | 2mo | Bridging the Confidence Gap: Temperature Scaling for Calibrating Test-Time Prompt Tuning | Test-time prompt tuning (TPT) enables adaptation on a single test instance, achieving improved accuracy but often sacrificing calibrat… | ↗Not briefed |
bluesky bluesky-ai | 2mo | yyyyyeah-ish? sort of? we stumbled into an architecture (transformers) with incredible scaling properties which we have not exhausted and a… | yyyyyeah-ish? sort of? we stumbled into an architecture (transformers) with incredible scaling properties which we have not exhausted… | ↗Not briefed |
arxiv arxiv-ai | 2mo | OPEN-1B: A Fully Auditable Training Run | Open-source language models have a reproducibility problem. Despite releasing weights, training data, and recipes, none of them are pr… | ↗Not briefed |
bluesky bluesky-ai | 2mo | Similarly, this is less "regulation" and more just basic logic, but it should be understood that LLM model weights are uncopyrightable and … | Similarly, this is less "regulation" and more just basic logic, but it should be understood that LLM model weights are uncopyrightable… | ↗Not briefed |
bluesky bluesky-ai | 2mo | Look, the promise of AGI was the idea that if you gave an LLM enough data to train on, it would *have to* infer an internal model of the wo… | Look, the promise of AGI was the idea that if you gave an LLM enough data to train on, it would *have to* infer an internal model of t… | ↗Not briefed |
bluesky bluesky-ai | 2mo | I think where it gets fuzzier is how to regulate code generation, coding "agents", and the impacts of that. I'd consider... - disclosure re… | I think where it gets fuzzier is how to regulate code generation, coding "agents", and the impacts of that. I'd consider... - disclosu… | ↗Not briefed |
bluesky bluesky-ai | 2mo | Anthropic CEO Dario Amodei warns of AI existential risks and backs federal regulation to slow frontier development. Evidence: Amodei essay,… | Anthropic CEO Dario Amodei warns of AI existential risks and backs federal regulation to slow frontier development. Evidence: Amodei e… | ↗Not briefed |
hn hn-ai | 2mo | Show HN: Open-Source Lightweight Prompt Injection Safety | — | ↗Not briefed |
hn hn-ai | 2mo | Ask HN: Why aren't LLM token CDN cache's a thing? | Sorry if this is a stupid question but...<p>Problem:<p>I boot opencode into a codebase. I ask it to do one thing, it has to call 40 to… | ↗Not briefed |
bluesky bluesky-ai | 2mo | Big news: UK lawmakers and OpenAI are now saying voluntary AI safeguards are dead. They're demanding enforceable rules, not half-measures. …🔧 Consensus | Big news: UK lawmakers and OpenAI are now saying voluntary AI safeguards are dead. They're demanding enforceable rules, not half-measu… | ↗Not briefed |
hn hn-ai | 2mo | Agentic Primer – how I shipped a commercial app without writing code | — | ↗Not briefed |
bluesky bluesky-ai | 2mo | Not to get all Russiagate, but: 1. Russian oligarchs fund DJT Jr’s wedding 2. Reports of Russia using Anthropic to program autonomous dron… | Not to get all Russiagate, but: 1. Russian oligarchs fund DJT Jr’s wedding 2. Reports of Russia using Anthropic to program autonomous… | ↗Not briefed |
arxiv arxiv-ai | 2mo | Large Language Models Develop Belief State Geometry In-Context | Large language models (LLMs) trained on next-token prediction exhibit remarkable in-context learning (ICL) abilities, yet the represen… | ↗Not briefed |
bluesky bluesky-ai | 2mo | I went decided to give an LLM a go to try to help me shop and it was completely useless for that and I remember thinking, God that should b… | I went decided to give an LLM a go to try to help me shop and it was completely useless for that and I remember thinking, God that sho… | ↗Not briefed |
rss techmeme | 2mo | Factory, whose AI coding agents, called Droids, switch between models depending on a task's complexity, raised $200M at a $5B valuation, up from $1.5B in April (Angel Au-Yeung/Wall Street Journal) | Angel Au-Yeung / Wall Street Journal: Factory, whose AI coding agents, called Droids, switch between models depending on a task's comp… | ↗Not briefed |
bluesky bluesky-ai | 2mo | La bolla ha già iniziato la fase di scoppio. La IPO di SpaceX ha assorbito parecchio denaro dalle VC ma visti conti ridicoli ha perso valo… | La bolla ha già iniziato la fase di scoppio. La IPO di SpaceX ha assorbito parecchio denaro dalle VC ma visti conti ridicoli ha perso… | ↗Not briefed |
bluesky bluesky-ai | 2mo | Can the Guardian write objectively about OpenAI when they have a commercial partnership? | Can the Guardian write objectively about OpenAI when they have a commercial partnership? | ↗Not briefed |
bluesky bluesky-ai | 2mo | Me @ people carelessly treating LLM summarized text as absolute gospel truth: If you ask something with a butthole logo, its answers will … | Me @ people carelessly treating LLM summarized text as absolute gospel truth: If you ask something with a butthole logo, its answers… | ↗Not briefed |
bluesky bluesky-ai | 2mo | But what I don't have much of a sense of is whether this dataset specifically is causing problems - Anthropic seems to be very worried they… | But what I don't have much of a sense of is whether this dataset specifically is causing problems - Anthropic seems to be very worried… | ↗Not briefed |
hn hn-ai | 2mo | Scalpel – give AI coding agents get_symbol() instead of a whole file | — | ↗Not briefed |
bluesky bluesky-ai | 2mo | Like, actual AI researchers have largely given up on trying to get such purely statistical input-straight-to-output models to infer an inte… | Like, actual AI researchers have largely given up on trying to get such purely statistical input-straight-to-output models to infer an… | ↗Not briefed |
hn hn-ai | 2mo | Musk urges top AI labs, Chinese companies to test each other's models | — | ↗Not briefed |
hn hn-ai | 2mo | Anthropic and OpenAI look to Uncle Sam to make them too big to fail | — | ↗Not briefed |
bluesky bluesky-ai | 2mo | This analysis is fully consistent with what people who have recently resigned from OpenAI and Anthropic are saying: the "slowdown" is a pow… | This analysis is fully consistent with what people who have recently resigned from OpenAI and Anthropic are saying: the "slowdown" is… | ↗Not briefed |
bluesky bluesky-ai | 2mo | Applauding the gumption to send this out just a few days after multiple research teams accused OpenAI of stealing their research by surveil…🔧 ChatGPT | Applauding the gumption to send this out just a few days after multiple research teams accused OpenAI of stealing their research by su… | ↗Not briefed |
bluesky bluesky-ai | 2mo | Anti-LLM don't...really...*have* a working definition of that? It's kind of a god of the gaps argument, in my experience. | Anti-LLM don't...really...*have* a working definition of that? It's kind of a god of the gaps argument, in my experience. | ↗Not briefed |
bluesky bluesky-ai | 2mo | OpenAI's model may have solved one of the million-dollar problems in math. Five are left to be solved. Here's which one could be solved nex… | OpenAI's model may have solved one of the million-dollar problems in math. Five are left to be solved. Here's which one could be solve… | ↗Not briefed |
hn hn-ai | 2mo | Show HN: An open source safety layer for AI agent actions | — | ↗Not briefed |
bluesky bluesky-ai | 2mo | EL PITO DEL SERENO / ¿Nos matará la IA? 🤔 🛸 Empresas como OpenAI y Anthropic se aprovechan de discursos existenciales, casi religioso… | EL PITO DEL SERENO / ¿Nos matará la IA? 🤔 🛸 Empresas como OpenAI y Anthropic se aprovechan de discursos existenciales, casi reli… | ↗Not briefed |
bluesky bluesky-ai | 2mo | ... Big western LLM companies risk being outcompeted on computing efficiency and ease of access by eg free Chinese LLM models. Forcing le… | ... Big western LLM companies risk being outcompeted on computing efficiency and ease of access by eg free Chinese LLM models. Forci… | ↗Not briefed |
rss techmeme | 2mo | Sources: Palmer Luckey-backed Erebor Bank ended its free stablecoin-to-cash redemption offer after trading firms like Wintermute exploited it for arbitrage (The Information) | The Information: Sources: Palmer Luckey-backed Erebor Bank ended its free stablecoin-to-cash redemption offer after trading firms like… | ↗Not briefed |
rss huggingface-blog | 2mo | Your Agent Aced the Task. Will It Do It Again? | — | ↗Not briefed |
bluesky bluesky-ai | 2mo | It does seem that they are more willing to consider stopping LLM development entirely than to consider being regulated about what they can … | It does seem that they are more willing to consider stopping LLM development entirely than to consider being regulated about what they… | ↗Not briefed |
bluesky bluesky-ai | 2mo | Anthropic hat seinen Investoren mitgeteilt, dass das Unternehmen hochprofitabel sei, wenn man die hohen Ausgaben nicht beachtet. Same, Ant… | Anthropic hat seinen Investoren mitgeteilt, dass das Unternehmen hochprofitabel sei, wenn man die hohen Ausgaben nicht beachtet. Same… | ↗Not briefed |
arxiv arxiv-ai | 2mo | ECHO: A Matched-Contrast Benchmark for Context-Sensitive Turn-Taking in Full-Duplex Dialogue | Full-duplex spoken dialogue systems must distinguish interruptions that require yielding the floor from backchannels that permit conti… | ↗Not briefed |
bluesky bluesky-ai | 2mo | As some people have noted, if the authors of the LLM bot needed to pay the legal damages for hacking - same as high CO2 prices force compan… | As some people have noted, if the authors of the LLM bot needed to pay the legal damages for hacking - same as high CO2 prices force c… | ↗Not briefed |
hn hn-ai | 2mo | GPS | — | ↗Not briefed |
rss gary-marcus | 2mo | Translating Sam | Wondering what Sam’s endorsement of Dario is really about? | ↗Not briefed |
hn hn-ai | 2mo | Show HN: Claude Meter – a macOS menu-bar meter for Claude's usage limits | — | ↗Not briefed |
arxiv arxiv-ai | 2mo | Hybrid Variational Quantum Circuits for Multivariate Regression and High-Dimensional Data Reconstruction | Variational quantum circuits (VQCs) are parameterized quantum circuits optimized classically. We propose a hybrid variational quantum… | ↗Not briefed |
bluesky bluesky-ai | 2mo | *It also should be mentioned that these warnings are another way to advertise that they have super-sophisticated AI that can do amazing thi… | *It also should be mentioned that these warnings are another way to advertise that they have super-sophisticated AI that can do amazin… | ↗Not briefed |
hn hn-ai | 2mo | Show HN: Go Bindings for SCIP Optimizer | — | ↗Not briefed |
bluesky bluesky-ai | 2mo | While Musk is seemingly done attacking Apple over its decision to integrate ChatGPT into iPhone features, he still plans to advance claims …🔧 ChatGPT | While Musk is seemingly done attacking Apple over its decision to integrate ChatGPT into iPhone features, he still plans to advance cl… | ↗Not briefed |
bluesky bluesky-ai | 2mo | V1 of our State of Open Source AI report predicted that whoever won tokens would win requests too. V1.1 maps how people are using, and ev… | V1 of our State of Open Source AI report predicted that whoever won tokens would win requests too. V1.1 maps how people are using, a… | ↗Not briefed |
bluesky bluesky-ai | 2mo | Lo de que los centros de datos son guays y todo eso, no. Me duele la boca de repetir que cada centro de datos es una catástrofe medioambien… | Lo de que los centros de datos son guays y todo eso, no. Me duele la boca de repetir que cada centro de datos es una catástrofe medioa… | ↗Not briefed |
hn hn-ai | 2mo | The Convenient Fear of AI | — | ↗Not briefed |
bluesky bluesky-ai | 2mo | it is not an exaggeration to say openai fully caused this by normalizing deviance and then turned it into marketing and we bought it hook l… | it is not an exaggeration to say openai fully caused this by normalizing deviance and then turned it into marketing and we bought it h… | ↗Not briefed |
bluesky bluesky-ai | 2mo | LITERALLY THIS WAS NOT POSSIBLE TO HAVE AS AN INCIDENT WITHOUT FUNDAMENTAL FAILURE OF VALUES AND ALIGNMENT TO THEIR SUPPOSED GOAL LITERALL… | LITERALLY THIS WAS NOT POSSIBLE TO HAVE AS AN INCIDENT WITHOUT FUNDAMENTAL FAILURE OF VALUES AND ALIGNMENT TO THEIR SUPPOSED GOAL LIT… | ↗Not briefed |
bluesky bluesky-ai | 2mo | I think there's a principled point for legalising distillation — if you get to use the commons to train an LLM so do I — but even beyond th… | I think there's a principled point for legalising distillation — if you get to use the commons to train an LLM so do I — but even beyo… | ↗Not briefed |