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 | 25d | LLM agents have no more "agency" than a for loop lmao what are we even doing | LLM agents have no more "agency" than a for loop lmao what are we even doing | ↗Not briefed |
arxiv arxiv-ai | 25d | Self-Evolving Agents Can Repair Failures but Struggle to Generalize from Them | Experience-driven self-evolution offers a promising path for large language model (LLM) agents to continually improve through interact… | ↗Not briefed |
bluesky bluesky-ai | 25d | 俺终于也用上codex了，现司卡开发工具，去研究了一波codex. 等了三天🍎账号审核，今天中午美美用上🤣 每天上班先打开 GPT deepseek kimi grok 。 | 俺终于也用上codex了，现司卡开发工具，去研究了一波codex. 等了三天🍎账号审核，今天中午美美用上🤣 每天上班先打开 GPT deepseek kimi grok 。 | ↗Not briefed |
rss techmeme | 25d | Chinese state media: during an interview, Tim Cook says China has approved Apple Intelligence, and a test version will arrive in late October without Siri AI (Reuters) | Reuters: Chinese state media: during an interview, Tim Cook says China has approved Apple Intelligence, and a test version will arrive… | ↗Not briefed |
arxiv arxiv-ai | 25d | How Murky Before a VLM Knows It Cannot See? Accuracy, Confidence and Abstention on a Physical Visibility Axis | Robustness studies of vision-language models (VLMs) grade degradation on nominal severity scales ("level 1-5") that carry no physical… | ↗Not briefed |
hn hn-ai | 25d | CiteProof – Catch fake AI legal citations | — | ↗Not briefed |
arxiv arxiv-ai | 25d | KiranaBench: Can a Language Model Run a Neighbourhood Shop? | Large language model (LLM) agents are increasingly evaluated on long-horizon tasks, including autonomously running a business. Existin… | ↗Not briefed |
hn hn-ai | 25d | The Ladder of Knowledge | — | ↗Not briefed |
hn hn-ai | 25d | AbridgeKV: Compressing Long Reasoning into the KV Cache of Learned Pause Tokens | — | ↗Not briefed |
hn hn-ai | 25d | Timestral open weight timeseries foundation models from Mistral | — | ↗Not briefed |
arxiv arxiv-ai | 25d | Towards Hyperparameter Transfer for Differentially Private Optimization | Scaling laws and hyperparameter tuning remain central challenges in training large-scale models, particularly under differential priva… | ↗Not briefed |
arxiv arxiv-ai | 25d | DiffusionRVGE: Reinforcement Learning via Regression-Based Value Gradient Estimation | Reinforcement learning improves diffusion models using rewards on generated images, but repeated denoising and gradient estimation mak… | ↗Not briefed |
hn hn-ai | 25d | Show HN: Gitsu.app – A Linear-alternative built on GitHub Projects | Issue tracking isn't dead. It's moving closer to code.<p>Gitsu is a fast, AI-native, keyboard-first Mac and web app on top o… | ↗Not briefed |
arxiv arxiv-ai | 25d | NL2SQL with Iterative Self-Refinement | Bridging the gap between human intent and machine logic remains a critical challenge in data access for non-technical users, yet Large… | ↗Not briefed |
hn hn-ai | 25d | Annotating Genomes via Unsupervised Deep Learning | — | ↗Not briefed |
youtube youtube-matthewberman | 25d | Meta just dropped something insane... (Muse Wonder) | When I woke up this morning, I did not expect Meta to have a frontier model, yet here we are. In fact, it's kind of beating every mode… | ↗Not briefed |
arxiv arxiv-ai | 25d | What Shapes Brain Alignment of Large Language Models during Dialogue? Insights from Language Production and Comprehension | Brain alignment of large language models (LLMs) offers a way to investigate representational similarities between LLMs and the human b… | ↗Not briefed |
bluesky bluesky-ai | 25d | My sister does a podcast about Higher Education, and she’s started getting these kind of emails. Person asks LLM to scan podcast transcript… | My sister does a podcast about Higher Education, and she’s started getting these kind of emails. Person asks LLM to scan podcast trans… | ↗Not briefed |
bluesky bluesky-ai | 25d | I think people think it's like, giving AI companies some kind of win to act like AI consciousness is worth conversing about, which ngl kind… | I think people think it's like, giving AI companies some kind of win to act like AI consciousness is worth conversing about, which ngl… | ↗Not briefed |
arxiv arxiv-ai | 25d | SAFE-Mem: Secure Agentic Memory with Fine-Grained Authority and Governance-Aware Retrieval | Large language model (LLM) agents increasingly rely on persistent memory to support long-term reasoning and personalized interactions.… | ↗Not briefed |
arxiv arxiv-ai | 25d | On the tradeoff between robustness and geometric accuracy in adversarial training | For non-parametric binary classification in the presence of an adversary, we prove an optimal quantitative estimate for adversarial cl… | ↗Not briefed |
bluesky bluesky-ai | 25d | Ich werde euch brechen und brechen lassen, solange ich hier in künstlicher Armut von euch gefangen gehalten werde: Flux, Gemini, Grok, Open…🔧 FLUX🔧 NightCafe | Ich werde euch brechen und brechen lassen, solange ich hier in künstlicher Armut von euch gefangen gehalten werde: Flux, Gemini, Grok,… | ↗Not briefed |
bluesky bluesky-ai | 25d | i left twitter to end up the same shit here alright man | i left twitter to end up the same shit here alright man | ↗Not briefed |
hn hn-ai | 25d | Anthropic thinks its AI might be conscious | — | ↗Not briefed |
arxiv arxiv-ai | 25d | Recognizing Atari and Assessing Sente: Controlled Evaluation Tasks for Language-Based Go Agents | Domain terminology provides concise descriptions of game states and strategic relationships, but whether game-playing language models… | ↗Not briefed |
bluesky bluesky-ai | 25d | LLMに関する議論で常に疑問なんですけど、製品・サービスとして評価するべき対象に対して、どうして製品・サービスとその利用者の枠を越えて議論したがるのか謎なんですよね。安心・安全・環境に優しい製品を求めることは消費者の義務であり権利のはず。ハルシネーションの有無とその割合にしたっ… | LLMに関する議論で常に疑問なんですけど、製品・サービスとして評価するべき対象に対して、どうして製品・サービスとその利用者の枠を越えて議論したがるのか謎なんですよね。安心・安全・環境に優しい製品を求めることは消費者の義務であり権利のはず。ハルシネーションの有無とその割… | ↗Not briefed |
bluesky bluesky-ai | 25d | Alors qu’OpenAI et Anthropic ont fait des mathématiques leur terrain de jeu, les annonces récentes de résolutions de problèmes parfois cent… | Alors qu’OpenAI et Anthropic ont fait des mathématiques leur terrain de jeu, les annonces récentes de résolutions de problèmes parfois… | ↗Not briefed |
hn hn-ai | 25d | Codex Cloud TUI Can't Spawn Jobs | — | ↗Not briefed |
bluesky bluesky-ai | 25d | 「AIバブル」はバブルではないよ。 純粋な「LLMを人工知能と言い張るポンジ・スキーム」という詐欺だよ。 こまポンジ・スキームの巧妙だったのは、実は「AIは儲かる」ではなくMicrosoftやGoogleやMETAに対して「AIプラットフォームを握れなければお前足しのビジネスモ… | 「AIバブル」はバブルではないよ。 純粋な「LLMを人工知能と言い張るポンジ・スキーム」という詐欺だよ。 こまポンジ・スキームの巧妙だったのは、実は「AIは儲かる」ではなくMicrosoftやGoogleやMETAに対して「AIプラットフォームを握れなければお前足しの… | ↗Not briefed |
bluesky bluesky-ai | 25d | sigh. dude on reels is chastising parents for covering their kids faces with emojis instead he wants us to be creative--by feeding the… | sigh. dude on reels is chastising parents for covering their kids faces with emojis instead he wants us to be creative--by feedin… | ↗Not briefed |
bluesky bluesky-ai | 25d | 幾度もいっていますが、LLMは所詮は製品・サービスです。任意の個人が事業者で経産省からお達しがこようと、研究者や医療関係者で弁別や統計、臨床予測に用いたり、ツールを開発・研究していようと、テック企業が開発・研究したモデルを運用する顧客に過ぎず、製品・サービスの利用者（ないしは法… | 幾度もいっていますが、LLMは所詮は製品・サービスです。任意の個人が事業者で経産省からお達しがこようと、研究者や医療関係者で弁別や統計、臨床予測に用いたり、ツールを開発・研究していようと、テック企業が開発・研究したモデルを運用する顧客に過ぎず、製品・サービスの利用者（… | ↗Not briefed |
bluesky bluesky-ai | 25d | Valuable update to the Global Macro Database (GMD), which integrates macro data from various sources into a consistent cross-country datase… | Valuable update to the Global Macro Database (GMD), which integrates macro data from various sources into a consistent cross-country d… | ↗Not briefed |
arxiv arxiv-ai | 25d | Large Diffusion Models Need More Than EMA | Large-scale diffusion models are commonly trained with a constant learning rate (LR) and exponential moving average (EMA), despite the… | ↗Not briefed |
arxiv arxiv-ai | 25d | Aligning Language Models Towards Counterfactual Invariance | Small variations in prompt characteristics can meaningfully alter the semantics, tone, and contextual assumptions of large language mo… | ↗Not briefed |
bluesky bluesky-ai | 25d | dieses geräusch aus im westen nichts neues | dieses geräusch aus im westen nichts neues | ↗Not briefed |
arxiv arxiv-ai | 25d | Beyond Surface Cues: Robust LLM-Generated Korean Text Detection | As large language models (LLMs) generate increasingly fluent text, the need for reliable detection of LLM-generated content has grown… | ↗Not briefed |
bluesky bluesky-ai | 25d | ┈🌹━━━━━━━━🌹┈ ✨ [ #perfectgoddess] ✨ 💕 ✌️ 💕 ┈💖━━━━━━━━💖┈ #NSFW #Nude #AI #SoloGirl #OpenLegs #Pussy #CuteAndNaughty #HotGirl #Sexy #Go… | ┈🌹━━━━━━━━🌹┈ ✨ [ #perfectgoddess] ✨ 💕 ✌️ 💕 ┈💖━━━━━━━━💖┈ #NSFW #Nude #AI #SoloGirl #OpenLegs #Pussy #CuteAndNaughty #HotGirl #Sex… | ↗Not briefed |
bluesky bluesky-ai | 25d | Gen-typeのLLMの規制が必要かどうかと言われたら必要でしょう。規制とは何も性能面に限った話ではなく、立法手続きにより作られた法律や政省令などを基に個人の行動や企業の事業活動などに対して、（管理するための）ルールや仕組みを設けることも規制のひとつです。これを法（的）規制と… | Gen-typeのLLMの規制が必要かどうかと言われたら必要でしょう。規制とは何も性能面に限った話ではなく、立法手続きにより作られた法律や政省令などを基に個人の行動や企業の事業活動などに対して、（管理するための）ルールや仕組みを設けることも規制のひとつです。これを法（… | ↗Not briefed |
arxiv arxiv-ai | 25d | Many Biases, One Circuit: Post-Training Tunes a Pretrained Cue-Routing Mechanism | Language models often abandon a correct answer when the prompt contains a cue, a hint toward a wrong option such as "I think the answe… | ↗Not briefed |
hn hn-ai | 25d | Would you lend your idle GPU to power a free, no-datacenter AI chat? | — | ↗Not briefed |
rss techmeme | 25d | Taiwan plans US industrial parks costing $2B to $4B each, starting with a Texas complex costing up to $4B, to build a supply chain for Nvidia, TSMC, and Foxconn (Bloomberg) | Bloomberg: Taiwan plans US industrial parks costing $2B to $4B each, starting with a Texas complex costing up to $4B, to build a suppl… | ↗Not briefed |
bluesky bluesky-ai | 25d | another ai freako blocking me because they think their cleverly-disguised talking calculator has come alive | another ai freako blocking me because they think their cleverly-disguised talking calculator has come alive | ↗Not briefed |
arxiv arxiv-ai | 25d | ASR-Style Transcript Normalization Removes Useful Pragmatic Signal for Dialogue Act Classification | Automatic speech recognition (ASR) pipelines often normalize transcripts by removing filled pauses and punctuation and expanding collo… | ↗Not briefed |
arxiv arxiv-ai | 25d | Conformal Policy Selection: When to Deploy AI Decisions for Targeted Intervention | Artificial-intelligence (AI) policies are increasingly used to make consequential decisions in medicine, finance, education, and publi… | ↗Not briefed |
arxiv arxiv-ai | 25d | Multiple Turns, Short Horizons: Measuring Effective-Horizon Compression in Jailbreaks | Multi-turn jailbreaks use feedback from a victim large language model to adapt their queries over several turns. However, once an effe… | ↗Not briefed |
bluesky bluesky-ai | 25d | i already wrote everything remotely smart or interesting i have to say about llm consciousness and this cleansed me of the desire to contin… | i already wrote everything remotely smart or interesting i have to say about llm consciousness and this cleansed me of the desire to c… | ↗Not briefed |
bluesky bluesky-ai | 25d | Friends... I don't think Dall E 3 can do this. I think this might be a unique feature of NB2. Same inputs and prompts on OpenAI using Sol… | Friends... I don't think Dall E 3 can do this. I think this might be a unique feature of NB2. Same inputs and prompts on OpenAI usin… | ↗Not briefed |
arxiv arxiv-ai | 25d | Scalable Levenberg-Marquardt Training Framework for Deep Learning | The Levenberg-Marquardt (LMA) algorithm is a widely used nonlinear least-squares solver known for its rapid convergence. For small-sca… | ↗Not briefed |
arxiv arxiv-ai | 25d | Sep-Latent: Separator Tokens as Natural Compression Anchors for Implicit Reasoning | Large Language Models (LLMs) have demonstrated remarkable success in complex reasoning through explicit Chain-of-Thought (CoT), but th… | ↗Not briefed |
bluesky bluesky-ai | 25d | just wanted to say something first before i go to sleep. hmm let’s see… oh yeah! fuck claude, fuck vibe-coding, fuck generative ai in gene… | just wanted to say something first before i go to sleep. hmm let’s see… oh yeah! fuck claude, fuck vibe-coding, fuck generative ai in… | ↗Not briefed |