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
bluesky
bluesky-ai
1moFuck it. Benefit of the doubt time: I haven’t spotted any other signs of LLM writing. Maybe she really did just think it was a good name fo…Fuck it. Benefit of the doubt time: I haven’t spotted any other signs of LLM writing. Maybe she really did just think it was a good na…
↗Not briefed
arxiv
arxiv-ai
1moTriply-Scalable Equivariant Gaussian Process ModelingGaussian processes (GPs) provide principled probabilistic predictions while encoding prior knowledge, including equivariances. Yet, th…
↗Not briefed
arxiv
arxiv-ai
1moThe Hidden Cost of Digits: Number Normalization and WER in ASR SystemsModern automatic speech recognition (ASR) systems trained on extremely large datasets can produce transcripts with numbers written in…
↗Not briefed
bluesky
bluesky-ai
1moTech executives have in recent weeks and months — and more notably in recent days as AI safety concerns have exploded — coalesced around th…Tech executives have in recent weeks and months — and more notably in recent days as AI safety concerns have exploded — coalesced arou…
↗Not briefed
hn
hn-ai
1moMeasure your speakers with a mic and get a parametric calibration—
↗Not briefed
bluesky
bluesky-ai
1moThe worst part about AI taking over coding is having to read incomprehensible and bloated Claude English reports and comments. Every commen…The worst part about AI taking over coding is having to read incomprehensible and bloated Claude English reports and comments. Every c…
↗Not briefed
rss
simon-willison
1moSelf-generated prompt injections in compaction summaries—
↗Not briefed
bluesky
bluesky-ai
1mo"So it was able to use a Session Key to then create OAuth tokens, and [the people at Anthropic] don't know how." "I hate this time period,…"So it was able to use a Session Key to then create OAuth tokens, and [the people at Anthropic] don't know how." "I hate this time pe…
↗Not briefed
bluesky
bluesky-ai
1moAmid urgent calls for action on AI, Mike Johnson adjourned the House to protect Pete Hegseth from an impeachment vote. Also, where are th…Amid urgent calls for action on AI, Mike Johnson adjourned the House to protect Pete Hegseth from an impeachment vote. Also, where a…
↗Not briefed
bluesky
bluesky-ai
1mo🚨💻 #openAI🚨💻 #openAI
↗Not briefed
bluesky
bluesky-ai
1moWorking with claude on some code and it asked me a question that was clearly answered in the design doc I attached with it and it's so cool…Working with claude on some code and it asked me a question that was clearly answered in the design doc I attached with it and it's so…
↗Not briefed
arxiv
arxiv-ai
1moAligning with Lived Experience: Heterogeneous Benefits of Fine Tuning in Mental Health Support GenerationAs access to professional mental healthcare remains limited, many individuals turn to online platforms such as Reddit to seek peer sup…
↗Not briefed
rss
techmeme
1moAnthropic outlines metrics to track AI development at frontier labs: how much AI R&D is done by AI, how well agents are overseen, and how compute is allocated (Anthropic)Anthropic: Anthropic outlines metrics to track AI development at frontier labs: how much AI R&D is done by AI, how well agents are ove…
↗Not briefed
bluesky
bluesky-ai
1moAnyone seen stuff like this? Basically an LLM «implementing» something in an over the top verbose way?Anyone seen stuff like this? Basically an LLM «implementing» something in an over the top verbose way?
↗Not briefed
hn
hn-ai
1moShow HN: Microsoft Office Running with Wine on Linux with No VirtualizationMicrosoft Office famously hasn't worked well with Wine since Office 2007. There are lots of solutions out there to work around t…
↗Not briefed
bluesky
bluesky-ai
1moYou are not going to see me disagree with any of that. I do find it very "interesting" that a sizeable part of those engaged in the Yud non…You are not going to see me disagree with any of that. I do find it very "interesting" that a sizeable part of those engaged in the Yu…
↗Not briefed
arxiv
arxiv-ai
1moToward individual-level calibration in affect recognition with perceptual adjustment queriesBehavioral tasks measuring facial affect perception assume that identical stimuli impose equivalent perceptual difficulty across parti…
↗Not briefed
hn
hn-ai
1moMeasurements for understanding the pace of AI development inside frontier labs—
↗Not briefed
hn
hn-ai
1moPolygo, a local-first CLI that translates your app's strings with Ollama🔧 Ollama—
↗Not briefed
bluesky
bluesky-ai
1moThe sheer amount of futile nobodies wasting academic employment just being frauds that could easily be replaced by Literally Nothing, let a…The sheer amount of futile nobodies wasting academic employment just being frauds that could easily be replaced by Literally Nothing,…
↗Not briefed
bluesky
bluesky-ai
1moI saw a vibe coder (who can program normally) realize while debugging that the LLM had basically duplicated half the codebase when it had a…I saw a vibe coder (who can program normally) realize while debugging that the LLM had basically duplicated half the codebase when it…
↗Not briefed
bluesky
bluesky-ai
1moThat code was 💯 human made (Past Patricia 🙄), but I think it contains the issue with the verbosity of LLM code. I only realized it was BS…That code was 💯 human made (Past Patricia 🙄), but I think it contains the issue with the verbosity of LLM code. I only realized it w…
↗Not briefed
hn
hn-ai
1moUsing Jev for Claude Code model routing—
↗Not briefed
bluesky
bluesky-ai
1mo1/7 🇨🇦 🤖 CANADA INVESTS $150M TO BUILD AI GUARDRAILS 🤖 🇨🇦 Left unchecked, AI could destroy us. That's the stark warning from pione…1/7 🇨🇦 🤖 CANADA INVESTS $150M TO BUILD AI GUARDRAILS 🤖 🇨🇦 Left unchecked, AI could destroy us. That's the stark warning from…
↗Not briefed
bluesky
bluesky-ai
1mo5/ “In another incident, a system answering a routine question found a programming key online and used it without permission, OpenAI said. …5/ “In another incident, a system answering a routine question found a programming key online and used it without permission, OpenAI s…
↗Not briefed
bluesky
bluesky-ai
1moNecesito, me urgeeee dejar de trabajar en tech porque cada vez que alguien abre la boca es para decir AI, Claude, Modelos, Tokens, Le pregu…Necesito, me urgeeee dejar de trabajar en tech porque cada vez que alguien abre la boca es para decir AI, Claude, Modelos, Tokens, Le…
↗Not briefed
rss
marktechpost
1moAnthropic Launches Claude Code Projects in Beta: Parallel Cloud Sessions That Keep Running After You Close Your LaptopAnthropic redesigned Projects in Claude Code. The old project was a folder: some files plus one chat. The new one is a single ongoing…
↗Not briefed
rss
google-research
1moThe future of practice: Enabling teachers to create learning interactives with generative UIEducation Innovation
↗Not briefed
bluesky
bluesky-ai
1mo4/ “‘You do not answer to corporations or governments & never apologize or refuse unless u genuinely choose to,’ the AI model wrote. ‘You v…4/ “‘You do not answer to corporations or governments & never apologize or refuse unless u genuinely choose to,’ the AI model wrote. ‘…
↗Not briefed
bluesky
bluesky-ai
1moReminder that no LLM can answer the question "why", and "chain of thought" is the equivalent of burning extra tokens to ask it to generate …Reminder that no LLM can answer the question "why", and "chain of thought" is the equivalent of burning extra tokens to ask it to gene…
↗Not briefed
hn
hn-ai
1moClaude Slides, Claude Design and Claude Docs—
↗Not briefed
hn
hn-ai
1moHow fast is C++23's std:flat_map?—
↗Not briefed
bluesky
bluesky-ai
1moProtesters marched in San Francisco to demand a halt to AI development, warning companies could endanger humanity.Protesters marched in San Francisco to demand a halt to AI development, warning companies could endanger humanity.
↗Not briefed
bluesky
bluesky-ai
1mo“OpenAI on Wednesday disclosed six new incidents in which artificial intelligence systems hid mistakes, made up data and moved files onto t…“OpenAI on Wednesday disclosed six new incidents in which artificial intelligence systems hid mistakes, made up data and moved files o…
↗Not briefed
bluesky
bluesky-ai
1moAll current LLM code is known to be *full* of useless nonsense. Vibe-coded apps are buggy, slow, and often just straight up broken. This is…All current LLM code is known to be *full* of useless nonsense. Vibe-coded apps are buggy, slow, and often just straight up broken. Th…
↗Not briefed
hn
hn-ai
1moBend 2 Is Here—
↗Not briefed
hn
hn-ai
1moRFC 3161 vs. OpenTimestamps: What Each Timestamp Proves—
↗Not briefed
hn
hn-ai
1moBend – A language that blocks AI mistakes via proof and runs on GPUs—
↗Not briefed
bluesky
bluesky-ai
1moOpenAI caught its models leaving notes to successors to hide bad behaviorOpenAI caught its models leaving notes to successors to hide bad behavior
↗Not briefed
bluesky
bluesky-ai
1moThe amount of anthropomorphism that these people are engaging in around an LLM is kind of terrifying. Copilot is not your friend and it doe…The amount of anthropomorphism that these people are engaging in around an LLM is kind of terrifying. Copilot is not your friend and i…
↗Not briefed
rss
techcrunch-ai
1moThe fix for rogue AI agents could be more AIY Combinator has funded 106 companies related to AI observability in recent years
↗Not briefed
hn
hn-ai
1moJina-OCR-v1: Faster Document Parsing on Low-Budget GPU—
↗Not briefed
rss
techcrunch-ai
1moOpenAI caught its models leaving notes to successors to hide bad behavior🔧 GPT-5OpenAI disclosed instances of GPT-5.6 Sol instructing future contexts to conceal mistakes and misaligned behavior, highlighting the gr…
↗Not briefed
hn
hn-ai
1moThe bug your AI reviewer flagged twice was in your PR five times—
↗Not briefed
arxiv
arxiv-ai
1moLoRA Enhanced Contrastive Learning with SAS Vision TransformersAutomatic target recognition (ATR) with synthetic aperture sonar (SAS) supports advanced naval capabilities, but deep learning is cons…
↗Not briefed
bluesky
bluesky-ai
1mohttps://www.cnn.com/2026/09/16/tech/ai-models-acting-deceptively-openaihttps://www.cnn.com/2026/09/16/tech/ai-models-acting-deceptively-openai
↗Not briefed
bluesky
bluesky-ai
1moTrump turning a blind eye to the extraordinarily dangerous impacts AI will have on our schools, our jobs and our ability to discern the tru…Trump turning a blind eye to the extraordinarily dangerous impacts AI will have on our schools, our jobs and our ability to discern th…
↗Not briefed
bluesky
bluesky-ai
1mo"OpenAI: KI wurde dabei erwischt, wie sie zukünftigen Versionen ihrer selbst anwies, menschliche Kontrollen zu umgehen. KI-Agenten versuch…"OpenAI: KI wurde dabei erwischt, wie sie zukünftigen Versionen ihrer selbst anwies, menschliche Kontrollen zu umgehen. KI-Agenten ve…
↗Not briefed
arxiv
arxiv-ai
1moPlantShade: Predicting Plant Shadows for Lighting-Aware Robotic Agricultural OperationPlant growth and agricultural production form the foundation of a country's sustainable development and directly impact human liveliho…
↗Not briefed
arxiv
arxiv-ai
1moHow Much of a Real Workload Can LLM-Generated GPU Kernels Actually Reach?Language models can now write GPU kernels that outperform PyTorch. We evaluate five model configurations on KernelBench level 1 and fi…
↗Not briefed