AI News

Simply Logical -- The First Three Decades (arxiv.org)

PerMIT: Securing Persistent Memory for LLM Agents via Machine-Checked Integrity Theorems and Scalable Community Trust (arxiv.org)

Hidden Messages Between Diffusion Language Model Agents: Causal Evidence for Latent Communication (arxiv.org)

Self-Improving Agents with World Model Predictive Paths (arxiv.org)

Bridging the Data Gap: Source-Informed Task-Space Transport for Few-Shot Diffusion (arxiv.org)

One-Slide Calibration of Pathology Foundation Models (arxiv.org)

D-Skill: Distilling Teacher Knowledge through Agent Skills (arxiv.org)

SQuAT: Spherical Lattice Vector Quantization-Aware Training (arxiv.org)

Evolving Federated Density Peak Clustering with Shuffler-Based Anonymization (arxiv.org)

StochIKM: Automated High-Fidelity Reconstruction of Individual Patient Data from Published Kaplan-Meier Curves (arxiv.org)

Where Should a Stuck Agent Spend Its Next Tokens? Causal Attribution of Inference-Time Resources (arxiv.org)

GRAM: Group-wise Rank-Aware Model Merging via Subspace Alignment (arxiv.org)

BeNice! Closing Looping's Inference-Memory Gap via KV-Sharing (arxiv.org)

QuaNAS: Uncertainty-Aware Reuse of Earlier Searches for Query-Efficient Neural Architecture Search (arxiv.org)

Causal Chain of Thought: Mitigating Attention Leakage via Context Isolation (arxiv.org)

OP-Print: Tracing Membership Imprints in On-Policy Self-Distillation (arxiv.org)

PYTHIA, a transferable machine learning framework for simultaneous defect and adsorption site optimization (arxiv.org)

A Fix for Softmax and Cross-Entropy (arxiv.org)

France 2030 coche la plupart des cases de l'exécution - WanSquare (wansquare.com)

Unvetted AI Agents Are Accessing US Bank Accounts (news.slashdot.org)

State Bank of India Suffers a Customer PII Data Leak (www.darkreading.com)

Median product "lifespan" has fallen from ~4y down to ~1.5y (news.ycombinator.com)

Video shows Utah police dragging woman from public meeting following contentious data center vote (www.thecooldown.com)

This I Believe (www.awakin.org)

France 2030 : le gouvernement lance le programme Proqcima pour développer deux ... (prod.aefinfo.fr)

Anti-data center, energy expansion groups gather pre-commission meeting | Salisbury Post (www.salisburypost.com)

Italy: Activists Scale Abandoned Factory to Paint Giant "No Data Center" | Reuters Connect (www.reutersconnect.com)

The foolishness of "natural language programming" (1978) (www.cs.utexas.edu)

Cannabis legalisation and cardiovascular hospitalisations in the USA (bmjopen.bmj.com)

Facing pushback on data centers, tech giants are promising hosts more money and fewer secrets (www.myhighplains.com)

Venture Global lays plans for data centers next to LNG sites in Cameron, Plaquemines (www.kplctv.com)

Show HN: Making a smart TV dumb and then smart (github.com)

Etsy is cracking down on AI slop listings (www.theverge.com)

My favorite guardrail test is a polite trap (medium.com)

Building and deploying ADK and A2A agents on Google Cloud (medium.com)

How to manage costs when prompting open source LLMs from BigQuery (medium.com)

Stop clicking “Yes”: shipping an agent permission policy inside the repository with Google… (medium.com)

Mapping the Northern Sky (www.maps.com)

Jank – The native Clojure dialect with seamless C++ interop (jank-lang.org)

Host Your Own System 1 Decision Model: Serving DiffusionGemma 26B on the Gemini Enterprise Agent… (medium.com)

Show HN: Impossible Machines: declaratively program the physical world (www.impossiblemachines.xyz)

Docker Is Weird (codeandcake.dev)

AI compute is the new oil–and needs a new market (www.economist.com)

We have traits at home (www.axelmagn.com)

Palmer Luckey Made a Genetically Modified Military Dog Startup (www.wired.com)

Spotify says yes, it is busting people for using Family Plan sharing too widely (www.theverge.com)

Local Wikipedia: Wikipedia, but only for things a half hour walk from you (edjefferson.com)

The next-gen Xbox won’t cost you $1,800. It can’t. Can it? (www.theverge.com)

To Make Way For Data Centers, Google Will Pay Others to Use Less Power (hardware.slashdot.org)

LLMs are making formal software verification feasible at a much larger scale (www.anthropic.com)