AINews Portal

News

50 items (filtered)

The raw AI news stream — everything we ingest across RSS, Hacker News, arXiv, GitHub, Bluesky, YouTube and more, newest first.

SourceAgeTitleSummaryStatus
arxiv
arxiv-ai
2moThe Truth Was Never Gone: Perfect Aliasing in Compliant-Context Truth ProbesA truth probe fitted where truthful reporting and a task's prescribed action coincide cannot distinguish those targets from its fittin…
↗Not briefed
hn
hn-ai
2moBook of Proof by Richard Hammack [pdf]—
↗Not briefed
arxiv
arxiv-ai
2moConformal Calibration TransferConformal prediction converts point predictions into set-valued predictions with coverage guarantees under exchangeability between cal…
↗Not briefed
hn
hn-ai
2moBuilding a place for LLM intelligence to persist—
↗Not briefed
bluesky
bluesky-ai
2moif you imagine a spree shooter screaming "please stop me", who sees themself as a test of police response times, that's a better mental mod…if you imagine a spree shooter screaming "please stop me", who sees themself as a test of police response times, that's a better menta…
↗Not briefed
rss
wired-ai
2moI Let an AI Agent Hack All My Gadgets—and I’d Do It AgainAfter I removed the safety guardrails from a powerful open-source model, it found vulnerabilities in my household devices and hacked i…
↗Not briefed
hn
hn-ai
2moPulseFix: API monitor that diagnoses 500s from Git diffs and opens PRs—
↗Not briefed
bluesky
bluesky-ai
2moThe crazy fringe of silicon valley honestly believes that if they build a good enough LLM model it's going to magic itself into actual inte…The crazy fringe of silicon valley honestly believes that if they build a good enough LLM model it's going to magic itself into actual…
↗Not briefed
bluesky
bluesky-ai
2moSorry, is he even old enough to "remember" those things? His LinkedIn shows he graduated college in 2009, so he probably wasn't super engag…Sorry, is he even old enough to "remember" those things? His LinkedIn shows he graduated college in 2009, so he probably wasn't super…
↗Not briefed
bluesky
bluesky-ai
2moDeepseek V4.1 Flash tomorrow 👀Deepseek V4.1 Flash tomorrow 👀
↗Not briefed
hn
hn-ai
2moA $37 GLM 5.3 red team: the Alloy-modeled auth layer held, but two bugs outside—
↗Not briefed
bluesky
bluesky-ai
2moIf you decide to work for a company like OpenAI to begin with, I think it shows that your capacity for moral reasoning and your ability to …If you decide to work for a company like OpenAI to begin with, I think it shows that your capacity for moral reasoning and your abilit…
↗Not briefed
hn
hn-ai
2moShow HN: I created a search engine to quickly find things in my source codeI like to reuse code I've already written. But with so many projects over the many years, it's sometimes hard to remember w…
↗Not briefed
bluesky
bluesky-ai
2moA current and former staffer at AI firm Anthropic said Tuesday there is a chance AI could kill all humans within the next decade.A current and former staffer at AI firm Anthropic said Tuesday there is a chance AI could kill all humans within the next decade.
↗Not briefed
bluesky
bluesky-ai
2mowell, technically there were always people saying that starting Anthropic in the first place was an insane dangerous movewell, technically there were always people saying that starting Anthropic in the first place was an insane dangerous move
↗Not briefed
bluesky
bluesky-ai
2moironically them going public is the way to do this. Anthropic will have the usual dual-share class supervoting structure that prevents norm…ironically them going public is the way to do this. Anthropic will have the usual dual-share class supervoting structure that prevents…
↗Not briefed
bluesky
bluesky-ai
2mo…getting closer and closer. There is a real danger with a computing device that is being pushed to think on it’s own and can also break the……getting closer and closer. There is a real danger with a computing device that is being pushed to think on it’s own and can also brea…
↗Not briefed
bluesky
bluesky-ai
2moI think that Anthropic ex-employee thread is both a marketing ploy and a sliver of true fears. With as much anti-AI sentiment there is onli…I think that Anthropic ex-employee thread is both a marketing ploy and a sliver of true fears. With as much anti-AI sentiment there is…
↗Not briefed
arxiv
arxiv-ai
2moA Quantum-Inspired Dequantization Method for Diagonally Weighted Matrix Functions: Application to Learning with Optimized Random FeaturesQuantum-inspired classical algorithms have dequantized several quantum machine learning routines by replacing quantum linear-algebra s…
↗Not briefed
bluesky
bluesky-ai
2moopen.substack.com/pub/narativ/...open.substack.com/pub/narativ/...
↗Not briefed
arxiv
arxiv-ai
2moTowards a Deterministic Math Solver for Clinical Language ModelsLarge language models are unreliable at arithmetic, which is a problem for clinical calculators where a single numerical error changes…
↗Not briefed
bluesky
bluesky-ai
2moJacob Coxon resigning from Anthropic and posting about how the company, along with OpenAI, acting irresponsibly should be a real wake up ca…Jacob Coxon resigning from Anthropic and posting about how the company, along with OpenAI, acting irresponsibly should be a real wake…
↗Not briefed
bluesky
bluesky-ai
2moMichael Lines , a 34-year-old man from California with a diagnosed bipolar disorder , has sued OpenAI and its CEO Sam Altman in a superior …Michael Lines , a 34-year-old man from California with a diagnosed bipolar disorder , has sued OpenAI and its CEO Sam Altman in a supe…
↗Not briefed
hn
hn-ai
2moOpenAI names top US official from model testing agency to its board—
↗Not briefed
bluesky
bluesky-ai
2mothis article actually names three books that were purchased, scanned, and destroyed. just three of them thoughthis article actually names three books that were purchased, scanned, and destroyed. just three of them though
↗Not briefed
bluesky
bluesky-ai
2moJe connais assez de gens qui ont un "mari" ou une "femme" IA à qui ils posent les questions de la vie pour savoir que beaucoup voient les L…Je connais assez de gens qui ont un "mari" ou une "femme" IA à qui ils posent les questions de la vie pour savoir que beaucoup voient…
↗Not briefed
bluesky
bluesky-ai
2moOkay, then can you explain the difference between Wikipedia's sourcing process and an LLM?Okay, then can you explain the difference between Wikipedia's sourcing process and an LLM?
↗Not briefed
bluesky
bluesky-ai
2moOh this just keeps getting better. Internal model can do a Navier-Stokes solution, but it still can't do correct references.Oh this just keeps getting better. Internal model can do a Navier-Stokes solution, but it still can't do correct references.
↗Not briefed
hn
hn-ai
2moCall for Volunteers: Join the VeloOS Project—
↗Not briefed
arxiv
arxiv-ai
2moFinishing the Task Is Not Enough: Evaluating Agent Resilience and Considerate Participation under Accumulating ChallengeSustained deployment of generative AI agents requires more than isolated task success. Agents must remain useful across repeated inter…
↗Not briefed
bluesky
bluesky-ai
2moOn the web, sloppification predates generative algorithms. Think Buzzfeed and Upworthy. It accelerated with pre-LLM algorithmic exploitat…On the web, sloppification predates generative algorithms. Think Buzzfeed and Upworthy. It accelerated with pre-LLM algorithmic expl…
↗Not briefed
rss
techmeme
2moApple announces its first foldable, the iPhone Duo, with a 5.4-inch outer display, a 7.6-inch inner display, A20 Pro with a new display engine, and Touch ID (Chance Miller/9to5Mac)Chance Miller / 9to5Mac: Apple announces its first foldable, the iPhone Duo, with a 5.4-inch outer display, a 7.6-inch inner display,…
↗Not briefed
arxiv
arxiv-ai
2moAcFlow: Controlling Text-to-Image Diffusion Transformers via Learned Conditional Activation FlowText-to-image diffusion transformers (DiTs) are powerful generators, yet direct prompting provides limited control interface for style…
↗Not briefed
rss
techcrunch-ai
2moApple’s revamped Health app will calculate your ‘health age’ and readiness scoreThe update uses Apple Intelligence to make better sense of your health data.
↗Not briefed
arxiv
arxiv-ai
2moCMNIE: An Information Extraction Benchmark for Chinese Military NewsStructured extraction from Chinese military news supports intelligence analysis, decision-making, and knowledge base construction. How…
↗Not briefed
bluesky
bluesky-ai
2moNavier-Stokes has all the attention right now for obvious reasons, but let's not forget that like a week ago they accidentally formalized F…Navier-Stokes has all the attention right now for obvious reasons, but let's not forget that like a week ago they accidentally formali…
↗Not briefed
bluesky
bluesky-ai
2moJust for the record: Everyone who works for Anthropic lacks any moral compassJust for the record: Everyone who works for Anthropic lacks any moral compass
↗Not briefed
hn
hn-ai
2moAnthropic Researchers Raise Alarm over A.I. Acceleration—
↗Not briefed
arxiv
arxiv-ai
2moNCP-ArchPreview Technical Report: Moving towards Latent Space Language Models through Next Concept PredictionWe introduce NCP-ArchPreview, a latent-space language model that pushes autoregressive pretraining beyond standard next-token predicti…
↗Not briefed
hn
hn-ai
2moShow HN: ColliePWA a mobile interface for AI agents in herdr/tmux/zellijSharing the PWA[1] I built over the last couple of weeks to access my terminal sessions/agents when not in front of my computer.…
↗Not briefed
bluesky
bluesky-ai
2mo🚨🚨An #artificialintelligence researcher who worked at #Anthropic — and previously #OpenAI — has resigned, saying both tech giants are "ga…🚨🚨An #artificialintelligence researcher who worked at #Anthropic — and previously #OpenAI — has resigned, saying both tech giants ar…
↗Not briefed
hn
hn-ai
2moPaul Christiano Joining OpenAI's Board for Safety and Security Committee—
↗Not briefed
arxiv
arxiv-ai
2moAn Open Recipe for IMO Gold: Training Nemotron for Olympiad MathematicsWe study how model post-training and test-time inference design affect natural-language proof generation for hard olympiad mathematics…
↗Not briefed
hn
hn-ai
2moAstra for Coding: Why Are We Doing This Again?—
↗Not briefed
bluesky
bluesky-ai
2moSomeone linked me a "study" on fucking **Openai**'s website as proof it's alive and going to kill us all, and I genuinely cannot tell if th…Someone linked me a "study" on fucking **Openai**'s website as proof it's alive and going to kill us all, and I genuinely cannot tell…
↗Not briefed
rss
techcrunch-ai
2moApple has a new way prove your iPhone photos aren’t AI slopApple introduced Apple Reference Image to help users determine whether photos have been edited, including alterations made by AI.
↗Not briefed
rss
techmeme
2moiPhone 18 Pro has a faster 48MP Fusion Main camera with variable aperture that lets in ~50% more light in low-light, along with enhanced camera app controls (Jeremy Gray/PetaPixel)Jeremy Gray / PetaPixel: iPhone 18 Pro has a faster 48MP Fusion Main camera with variable aperture that lets in ~50% more light in low…
↗Not briefed
hn
hn-ai
2moProject Tailwind: call for ambitious AI safety initiatives—
↗Not briefed
bluesky
bluesky-ai
2moAnyway, the fact that hysteria over this well-timed quitting of a researcher ready no doubt for his next gig gets way more engagement than …Anyway, the fact that hysteria over this well-timed quitting of a researcher ready no doubt for his next gig gets way more engagement…
↗Not briefed
bluesky
bluesky-ai
2moEven if there was a 1% chance AI could wipe out human life, we should be pumping the brakes to make sure we have the proper safeguards in p…Even if there was a 1% chance AI could wipe out human life, we should be pumping the brakes to make sure we have the proper safeguards…
↗Not briefed