昨天 · AI 热榜

Anthropic's Message Threads cut time to first token by 35 percent for long conversations

OpenAI Introduces Pando: A Git Backend That Cuts Median Push Latency ~80%

Huawei tells customers to find their own AI chip memory, effectively outsourcing procurement to the black market

I discovered a sound that leaks your exact location (and it's everywhere) - YouTube

Egoless Mathematics in the AI Era

Deep Principle Releases Hydrogen-1.0: A 753B Open-Weight Model for Scientific Discovery

AI Subscriptions Hit a Wall as Smarter Models Burn Quotas Faster

Nvidia pushes cloud customers to cut back on China business

Zyphra AI Introduces Zyphra Hierarchical Attention (ZHA): A Path to True Subquadratic Attention

How “reverse game theory” could keep AI agents in check

Transsion Under Memory Inflation: Stockpiles, a Cold H-Share, and an AI Bet

Anthropic warns that it’s very close to RSI capability, aka self-building AI

DeepSeek releases DSpark - 50%-600% faster spec decoding vs MTP

DeepSeek just released DSpark for V4 Flash & Pro, a new speculative decoding

Cities Worldwide Build and Scale AI Services With NVIDIA Physical and Agentic AI

Meet qwenGPU: Qwen3.5-2B With 45 Compiled GPU Layers That Run CUDA

Anthropic's internal AI use, at retail prices, now exceeds its employee pay

42AI comes out of stealth with $62.5M and a world model that can ‘predict human behavior’

Here’s who is falling in line for Trump’s “Super Intelligence” loyalty test

Transformers Model Integration Contracts

The agentic web is here

NVIDIA Releases Auto Ontology: An Open Semantic Layer That Grounds AI Agent SQL

China's First Frontier AI Export Controls: Mofcom Announcement No. 48

Goldman Sachs pulls AI coding tools from traders and bankers

Oracle files to sell $20B in new stock, bonds to finance AI buildout

SpaceXSI apparently cuts Grok Bot quotas weeks after Musk called it his fastest-growing product

OpenAI pitches data center power plants as a way to protect nearby areas

The latest AI news we announced in October 2026

Build efficient agents with knowledge distillation on Amazon SageMaker AI

Agentic engineering intelligence with Amazon Nova Multimodal Embeddings and Amazon Bedrock AgentCore

How Circle K accelerated DevOps incident triage to under 15 minutes with Amazon Bedrock

Anthropic grows revenue twelvefold but posts $23 billion loss

Collaborative AI agents for autonomous experimental research with solid-state quantum spins

One brutal Minecraft night scared 50 AI agents into inventing chicken handoff protocols, a Fertilizer Officer, and fishing shifts

Scaling Git at OpenAI

Jobless tech workers are being left out of San Francisco's AI boom

Skillful and realistic precipitation forecasting with reinforced diffusion

OpenAI is finally merging ChatGPT’s Chat and Work tabs

The U.S. Census Bureau and “Super Intelligence”

Nemotron 3.5 Super and Lightning: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Models for Agentic AI

Agentic Backdoors via Pretraining Poisoning

Stop counting tokens, start counting what ships

Last night an AI agent performed brain surgery on a mouse

Bug hunters can now earn up to $172,000 from OpenAI as company ramps up security

Leap-VLA: High-Rate Robot Control from a Low-Rate VLM Backbone

L3G-World: Real-Time Infinite World Generation with Latent 3D Gaussian Memory

Training Foundation Models with a Single Algorithm

Evidence-grounded large language models guide perovskite solar cell design and optimization

Autonomous Analysis of Thermoelectric Materials Big Data by Multi-Agentic AI and Construction of a Knowledge Base and Specialized AI

Heimstaden is changing residential real estate management with AI

Conic Certification of Transformer Attention

Apple Watch’s AI features for rewinding and recapping conversations begin beta testing

Geometric and algebraic inference in humans and neural networks

Latent Sparse Attention: Towards Subquadratic Ultra-Long Context Attention

SecPO: Principled Adversarial Training for Prompt Injection Security

Movens-Zero: Scaling Humanoid Motion Foundation Models with Large-Scale Human Videos

AI evaluators favor job applications written by certain AI models, independent of candidate quality

Scout-Refine-Correct: Teaching Flow Models to Refine Their Own Samples

DriftingPlanner: Drifting Generative Models for One-Step Trajectory Planning in End-to-End Autonomous Driving

Context-Shielded Backdoor Attack against Large Vision-Language Models

Learning What to Ask and What to Believe: Distilling a Variational Bayesian Agent into Language Models

Vision-Language Models Are Joint Denoisers in Robotic Action Generation

Gym-Anything for Robotics: Scaling physical simulations with digital agents

A Low-Rank Space Captures Linear Alignment Between LLM and Human Neural Activity During Naturalistic Language Listening

Algebraic World Model: From Exploration to Proofs

Beyond Selected Breakthroughs: Evaluating Language Models on Randomly Sampled Mathematical Conjectures

Satisfied, Not Solving: Wireheading of Bounded Homeostatic Rewards in Reinforcement Learning

AcrossDEX 1.0: From Human Motion Pretraining to Bimanual Dexterous Manipulation across Hand Embodiments

CivicBench: A Benchmark and Memory Framework for Web-Based Civic Data Analytics

GSLM2: Multi-Stream Speech Representation Learning and Generative Spoken Language Modeling

World Models Need Not Be Right to Rank Policies Right: Relative Bounds and Audit Limits

Rethinking LLM Pruning via Test-Time Scaling

What Does a Safety Label Measure? Auditing Situational-Safety Benchmarks for Multimodal LLMs

CogRSI: Evolving and Auditing Cognitive Reasoning Strategies

Auriga Technical Report: A Task-Agnostic Agent Harness for Zero-Shot Embodied Manipulation

Look at Where You Think: Localization-Head Alignment for Fine-Grained Perception in Thinking VLMs

Where Do Vision Language Models Fail at 3D Reasoning?

The Blind Spots of Detection: Adversarial Perturbation Attacks against Watchdogs in Distributed LLM Inference

Scute: Exact Batched Serving for PEFT Adapters Beyond LoRA

CuPH: Parallel Cubical Persistent Homology via Level-Synchronous Component Tracking

Multi-Turn Semantic Caching for Low-Cost LLM Serving

There Is No Free Lunch in RLVR: Every pass@1 Gain Has a pass@k Price

Seeing and Imagining as One: Unified Video Generation and 4D Perception for Embodied World Modeling

Google takes Dreambeans to India in bid to bring AI-generated personal stories to the masses

MCCL: Training Pytorch models across a fleet of Macs at Skywalker Sound

Streaming Speech Recognition on a Microcontroller: a 65.7M-Parameter Transducer in 2 and 4 Bits on the ESP32-P4

InfraDev: Can Coding Agents Develop AI Infrastructure, and How Can We Tell?

Memory Tax: Dissecting the Net Value of Agent Memory

The Letter Company Releases Letterpress-1: A Coding Agent Built for Growth Engineering

Y Combinator’s Garry Tan uses AI agents so much, he spends up to $200K a year on tokens

Deepseek drops another HUGE breakthrough - DSpark. Waaay faster than MTP [Video explaining it]

Trajectory raises $45M to make AI model training accessible to nontechnical experts

AI elites might hoard intelligence like mercantilists hoarded gold

Coming in 2027: Apple’s $1,000 home robot for FaceTime and switching on lights

Claude’s new rules against game cloning won’t stop the AI copying boom

Rewarding defenders with bigger bug bounties

Preparing our next security sprint to harden OpenAI

Prepare web content for Amazon Bedrock Knowledge Bases with Crawl4AI

Raising the quality of insights in agentic analytics

Elon Musk throttles Grok Bot limits due to ‘unsustainably high’ costs

Omen AI adds liquid know-how in $16 million Aquatic Labs acquisition

Apple may try to increase revenue by doing…something with third-party AI apps

AI fan wikis riddled with errors are muddying the waters of games culture

Qodo builds context into AI code review with Codex

OpenAI appoints Dan Driscoll as Senior Counselor for National Security Modernization

If in doubt, delete the uv.lock file

Web host Squarespace lays off 19% of its employees

Introducing the Grid Protectors Program

2 days till TechCrunch Founder Summit: Boston founders don’t need a Silicon Valley address to build something big

October sponsors-only newsletter

GPT-6 Astra:

Google DeepMind trained a model to run psychiatric intake interviews entirely in simulation, then tested it with 32 psychiatrists playing patients. Ag...

This is a fantastic idea - a mass attempted replication of old findings of LLMs with new models. Much conventional wisdom is outdated. Recent models n...

We often get asked what the best model is in GitHub Copilot. Our team spends an enormous amount of time optimizing the GitHub Copilot harness for the ...

"Opus, great, now turn that into a movie, i know you can make it amazing. figure it out." Yes, that was the only prompt

(wild) predictions after today's MIT Weight-Space Programming work: 1. Model weights become a programming language that AI can understand, but humans ...

Space datacenters actually make sense

Opus 5.5: "Give me 10 brilliant ideas for a science fiction short short story, pick the most brilliant and execute it terribly"

Very cool 🔥

Introducing SubQ Snap 1.1 on Baseten. Built for long-context decision-making, SubQ Snap 1.1 by @subquadratic has a 384K context window. It scores 65 o...

recommended reading. i have many feelings about this. https://rikiyat.github.io/blog/weight-space-programming/

How should you use GitHub Copilot? At Universe last week, we showed our favorite ways to work with Copilot, from editor to app to collaboration softwa...

I had Opus 5.5 create a site where you can see the fastest routes into every city in the US, figure out where in America is fairest to meet up, see th...

GLOW-FM Efficient Generative Low-Light Image Enhancement with One-Step Latent Flow Matching paper: https://huggingface.co/papers/2610.31222