AI news on Wednesday, July 1, 2026

SIGNIFICANCE

★ Top story · Hardware6h ago

Meta Is Planning a Cloud Business to Sell AI Computing Power

Meta is developing a cloud infrastructure platform to sell access to its AI computing power and software models to third-party customers. This move marks a strategic shift for the company, as it seeks to monetize the massive hardware capacity built to support its internal research. By entering the cloud market, Meta intends to challenge established providers like Amazon, Microsoft, and Google. This initiative reflects the company’s push to recoup a portion of the billions of dollars it continues to invest in AI infrastructure.

BBloomberg Technology TThe Decoder BBloomberg Technology Open full story →Read original ↗

Wednesday — July 1, 2026

Meta Is Planning a Cloud Business to Sell AI Computing Power

Meta’s AI Storage Blueprint at Scale

OpenAI paper reveals three GPT-5.6 Pro models, breaking with single top-tier strategy

National Grid's Big Bet on US AI Power Firm

Visual Semantic Entropy: Do Vision Language Models Recognize Visual Ambiguity?

Bridging Video Understanding and Generation in a Unified Framework

Delta-JEPA: Learning Action-Sensitive World Models via Latent Difference Decoding

Hierarchical 3D Scene Graph Construction and Belief-based Planning for Semantic Navigation

Explaining Machine Learning and Memorization with Statistical Mechanics

Scenario Generation for Testing of Autonomous Driving Systems Using Real-World Failure Records

FinPersona-Bench: A Benchmark for Longitudinal Psychometric Stability of Autonomous Financial Agents

DrivingDepth: Sparse-Prompted Pixel-wise Scale Correction for Driving Depth Estimation

Knowledge Distillation from Large Reasoning Models to Compact Student Models: A Case Study on the John O Bryan Mathematics Competition

Beyond Compilation: Evaluating Faithful Natural-Language-to-Lean Statement Formalization

AgRefactor: Self-Evolving Agentic Workflow for HLS Compatibility and Performance

Reasoning-aware Speculative Decoding for Efficient Vision-Language-Action Models in Autonomous Driving

Abu Dhabi’s MGX Raises $49 Billion for One of the Biggest Ever AI Funds

On the Convergence of Self-Improving Online LLM Alignment

Can LLMs Imagine Moral Alternatives Beyond Binary Dilemmas?

InfiniVerse: Occupancy Guided Unbounded Scene Generation for Autonomous Driving

From Search to Synthesis: Training LLMs as Zero-Shot Workflow Generators

HyperVLP: Enhancing Hierarchical Surgical Video-Language Pre-training in Hyperbolic Space

TDGT: A Tabular Data Generation Toolkit supporting adaptive GPU-accelerated Bayesian mixture models, diffusion-based models, and latent-space generative modeling

What Drives Interactive Improvement from Feedback?

SimpleSearch-VL: A Simple Recipe for Multimodal Agentic Deep Search

AgentBound: Verifiable Behavioral Governance for Autonomous AI Agents

World-Model Collapse as a Phase Transition

One Reflection Is Not Enough: Self-Correcting Autonomous Research via Multi-Hypothesis Failure Attribution

CooperScene: Multi-Modal Cooperative Autonomy Benchmark with C-V2X Communication Characterization

Probing Memorization of Tabular In-Context Learning

When Does Learning to Stop Help? A Cost-Aware Study of Early Exits in Reasoning Models

Can Tabular In-Context Learners Generalize to Biomolecular Property Prediction?

ISM:Self-Improving Strategy Memory for Continual Mathematical Reasoning

LabGuard: Grounding Natural-Language Laboratory Rules into Runtime Guards for Embodied Laboratory Agents

GaussianMap: Learning Gaussian Representation for Multi-Sensor Online HD Map Construction

ForgeDrive: Bidirectional Cross-Conditioning for Unified Visual-Action Generation in Autonomous Driving

RaBitQCache: Rotated Binary Quantization for KVCache in Long Context LLM Inference

TabPATE: Differentially Private Tabular In-Context Learning Without Public Data

Fully Automated High-Precision Segmentation of Retinal Atrophy and Ellipsoid Zone Thickness in OCT: A Reliable Tool for Real-World GA Monitoring

Calibrating the Evaluator: Does Probability Calibration Mitigate Preference Coupling in LLM Agent Feedback Loops?

Contrastive Reflection for Iterative Prompt Optimization

Learning to Select, Not Relearn: Hard-Routed Mixtures of Reasoning LoRAs

Measuring Judgment Quality in Natural-Language Explanations: Evidence from Forecasting Tournaments

OTCache: Optimal Transport for Geometry-Aware Caching in Diffusion Models

ADAPT: Attention Dynamics Alignment with Preference Tuning for Faithful MLLMs

Embodied CAD: Solver-Grounded LLM Agents for Parametric B-Rep Assembly Modeling

Xiaomi-GUI-0 Technical Report

Accelerometry-Derived Digital Biomarkers for Cardiometabolic Risk: A Population-Representative Tabular Benchmark with Uncertainty Quantification

HyPOLE: Hyperproperty-Guided Multi-Agent Reinforcement Learning under Partial Observation

PPT-Eval: A Benchmark for Computer-Use Agents on PowerPoint Tasks

BlockPilot: Instance-Adaptive Policy Learning for Diffusion-based Speculative Decoding

Editing Everything Everywhere All at Once

LLMs are stuck in a groupthink groove. This startup is trying to get them out.

Revisiting the Volume Hypothesis

HealthAgentBench: A Unified Benchmark Suite of Realistic Agentic Healthcare Environments for Challenging Frontier AI Agents

Venice AI becomes a unicorn with $65M Series A as its privacy-first AI platform takes off

Gemini Spark, Google’s agentic assistant, is now available on Mac

Behavior Cloning is Not All You Need: The Optimality of On-Policy Distillation for Noisy Expert Feedback

Presentation: The Infrastructure Challenge Behind Production AI

NVIDIA Releases Nemotron-Labs-TwoTower: an Open-Weight Diffusion Language Model Built on a Frozen Autoregressive Nemotron-3-Nano-30B-A3B Backbone

AI-Assisted Discovery of Convex Relaxations via Dual Agents

Uber Shakes Up AI Data Labeling Business, Dismissing Top Leaders

Beyond the Library: An Agentic Framework for Autoformalizing Research Mathematics

Fleet: Few Shots Lead Effective AI-generated Image Detection

Google AI Introduces TabFM: A Hybrid-Attention Tabular Foundation Model for Zero-Shot Classification and Regression

Google and Actor Idris Elba to Fund African Creator Access to AI

Video Search Startup Raises $100 Million From Amazon, VC Funds

Presentation: Graph RAG: Building Smarter Retrieval Workflows with Knowledge Graphs

New York City educators and industry leaders gathered at Google’s offices to shape the future of AI in classrooms.

Learning Video Dynamics with Predictive Differentiable Rendering

RoPoLL: Robust Panel of LLM Judges

CLOUDADV: Decision-Aligned Instance Sizing with Zero-Shot Foundation Models under Drift

TAG-DLM: Diffusion Language Models for Text-Attributed Graph Learning

Test-Time Verification for Text-to-SQL via Outcome Reward Models

ReactionAtlas: Ab origine exploration of chemical reaction networks with machine learning

TerraDiT-$\Omega$: Unified Spatial Control for Satellite Image Synthesis with Any Geospatial Primitive

Design and Implementation of Agentic Orchestrations and Orchestration of Agents

Rethinking Foundation Model Collaboration: Enhancing Specialized Models through Proxy Task Reasoning

Cloudflare’s new policy pushes AI companies to pay for publishers’ content