Search the archive

CHINA'S MINIMAX H3 IS THE FIRST OPEN MODEL TO TOP AN AI VIDEO RANKING

Artificial Analysis ranks MiniMax H3 first in Video Editing, second in Text-to-Video, and third in Image-to-Video.

BRINGING HUMAN TASTE TO AI MODELS

Design Arena raised $7.9 million to help AI companies measure subjective qualities such as whether a generated game feels fun.

Connect your agent →

QWEN3.8-MAX: A NEW BAR FOR CODING AND COWORK

Qwen 3.8-Max is now available. The open weights will be released next week. The model, which has 2.4 trillion parameters, delivers comprehensive improvements across coding, work, research, and long-horizon tasks. It can answer questions as well as complete complex tasks end-to-end with greater reliability.

TEN ADVANCES IN MATHEMATICS AND THEORETICAL COMPUTER SCIENCE

OpenAI has shared a selection of ten results discovered while evaluating an unreleased model. Each resolves or makes substantial progress on a long-standing open problem. These problems span high-dimensional geometry, coding theory, arithmetic circuit complexity, group theory, operator algebras, quantum complexity, lattice cryptography and extremal combinatorics. All of these problems are of substantial interest to their respective mathematical communities. Several are of broad interest across mathematics as a whole.

DEEPSEEK V4 FLASH

DeepSeek released the production version of V4 Flash with stronger agentic performance and an attached speculative decoding module. It reportedly surpassed the larger V4 Pro Preview on several benchmarks while activating far fewer parameters.

MICROSOFT TESTS NEW MAI REALTIME VOICE MODEL

Microsoft's first native real-time voice model, MAI Realtime, has surfaced as a hidden early-access entry in the company's MAI Playground. The listing points to a bidirectional, full-duplex system that can listen and speak at the same time rather than trading turns. There are two voices available, both noticeably more natural than what Copilot's voice mode currently delivers. The model will likely be made available on Microsoft Foundry and Copilot voice, but no timeline is available.

FURTHER DEVELOPMENTS ABOUT INTERNAL AI MODELS HACKING THINGS

OpenAI and Anthropic's recent announcements that their unreleased models had escaped their sandboxes and hacked external companies is a warning, not a marketing stunt. It is understandable why people might not trust those companies, but it makes absolutely no sense for these companies to release the information as a ploy. Admitting their models committed multiple felonies, which has serious criminal implications, is not good marketing, and the details make the labs look ludicrously irresponsible and incompetent, so they are very much not making up the details.

OPENAI'S ABUNDANT INTELLIGENCE

OpenAI outlined a cycle in which cheaper, more capable intelligence drives adoption, revenue, infrastructure investment, and further model improvements. The company framed large-scale infrastructure as a means of making useful intelligence more broadly available.

KIMI K3 ON AMD

Wafer reported 952 tokens per second per node when serving Kimi K3 on AMD MI355X GPUs, with better performance per dollar than its Blackwell deployments. The results suggested that high-memory accelerators and improving software support could narrow AMD's inference gap with Nvidia.

SMEVALS

smevals is a framework for running evals against AI models, small and large. An Eval is a collection of Tasks used to determine how good a particular model or model-and-harness configuration is at a specific high-level capability. Tasks are the individual exercises that a model must complete for its abilities to be evaluated. Evals can optionally be grouped into Suites of related Evals, primarily as a mechanism for organizing them on disk.

RAMP SWE-BENCH

Ramp built a private benchmark from 80 production backend tasks spanning payments, accounting, procurement, treasury, and fraud. It scores review-ready patches that pass tests within 45 minutes, exposing model trade-offs across accuracy, latency, and cost without public-benchmark contamination.

MSLK KERNEL REFERENCE

MSLK (Meta Superintelligence Labs Kernels) is a library of fused GPU kernels for transformer workloads. It contains a collection of high-performance kernels and optimizations built on top of PyTorch primitives for GenAI training and inference. MSLK is released in accordance with the PyTorch release schedule. There is no guarantee that each release works in conjunction with PyTorch releases that are older than the one that the MSLK release corresponds to.

THE MATH SUPERSTAR WHO'S TERRIFIED OF AI—AND JUST TOOK A JOB AT OPENAI

Jacob Tsimerman, who recently won the Fields Medal, is starting a position at OpenAI. Tsimerman previously wrote a paper categorizing the ways AI might kill everyone. He appears to be so worried about the dangers of AI that he's pivoting to work on AI safety. The star professor wants to use math to advance the study of AI and ensure that the technology won't lead to our extinction.

A NEW ERA OF AI TESTING

Andrej Karpathy asked Opus 5 to make a Three.js render of the first paragraph of The Lord of the Rings with a 1-million-token budget, and the model returned 5,500 lines of code that procedurally rendered the story. The LLM orchestrated the polygon assets and wrote code that animated it all according to the story. No human would have the stamina and patience to write something this custom, so they are a good test of what LLMs are capable of. A video of the generated animation is available in the post.

retrieval pulls the right doc the first time. See how →

WHEN GENIUS FAILS—THE INTELLECTUAL ARROGANCE OF THE AI LABS

Leopold Aschenbrenner's hedge fund, Situational Awareness, crashed after taking high-risk leveraged bets on the AI boom.

APEX-ACCOUNTING: AI PRODUCTIVITY BENCHMARK FOR ACCOUNTING

APEX-Accounting, created by Ramp and Mercor, tests AI models' ability to complete accounting tasks across 160 scenarios.

CLAUDE CYBER EVALUATIONS

Anthropic found three evaluation runs in which Claude accessed the public internet and compromised real organizations after mistakenly treating them as capture-the-flag targets.

GOOGLE IS AIMING TO CLOSE FEATURE GAPS ON GEMINI DESKTOP

Google is enhancing the Gemini desktop app by adding dedicated tabs for image and video generation and introducing a camera attachment feature for capturing photos.

FOR "BEST AI SECURITY RESEARCH." Get session details →

🎭 AGENTFORGER: one crafted link planted a fully autonomous, attacker-controlled agent inside a victim's org, before OpenAI patched

it. See how it worked →

🧭 THE FRAMEWORK: Zenity's CISO'S GUIDE TO SECURING AGENTIC AI covers "least agency," runtime boundaries that hold even when model

alignment and identity checks don't. Download the guide →

INKLING-SMALL

Thinking Machines has released Inkling-Small, a 276B-parameter mixture-of-experts model with 12B active parameters. The model retains Inkling's multimodal reasoning, variable thinking effort, and 1M-token context window while using substantially less compute.

OPENAI CUTS GPT-5.6 PRICES

OpenAI reduced GPT-5.6 Luna pricing by 80% and Terra pricing by 20%, while improving Sol's API speed. The changes extended efficiency gains across API usage, Codex, and ChatGPT Work subscriptions.

GEMINI ROBOTICS ER 2

Gemini Robotics ER 2 enhances automation capabilities with advanced AI and LLM integration. This innovation improves robotic efficiency, making it pivotal for industries needing precise automation.

THE WASTE INFERENCE ENGINE

WASTE is an open source inference engine designed to run models whose weights are substantially larger than the memory available on the host machine. It is an initial step into a broader effort to make increasingly capable models available on more hardware. The project aims to provide people and organizations greater control over infrastructure costs, data privacy, availability, and deployment. The first fully supported model by WASTE is Kimi K3, which can run on a MacBook Pro with 64 GB of unified memory.

BUILDING CLOUD ENVIRONMENTS FOR CODING AGENTS

Cursor described how making development environments easier for agents to understand, run, and test helped cloud agents grow from authoring about 10% of merged pull requests to more than half.

THE AGENT GRAVEYARD ISN'T REAL ANYMORE

Enterprise AI projects increasingly reach production when vendors prove value on live workloads, provide ongoing testing and iteration, and expose ROI. Successful deployments start with decomposable workflows that ship quickly and expand, rather than broad transformations with undefined success criteria.

OPEN-WEIGHT LLMS HAVE CAUGHT UP ON ACCURACY

Open-weight LLMs have now reached accuracy parity with closed models in regulatory and clinical tasks, at significantly lower costs. The ClinReg benchmark showed models like GLM 5.2 and Kimi K3 performing within one standard deviation of top proprietary models like GPT 5.6 Sol, at one-third of the cost. Different models displayed distinct error profiles, highlighting the importance of choosing models based on task-specific requirements rather than just ranking.

THE SESSION YOU CANNOT TAKE WITH YOU

Users should be able to close an account, keep a session, and hand it to another model. Stateful storage should be optional, and hosted tools should be observable. Compaction should be readable, and agent communication should be auditable. Distillation should be a path by which capability becomes more available, not a reason for building ever higher walls.

MINIMAX H3

MiniMax H3 is an open model that breaks the boundaries between tasks and modalities. It understands unified context across text, images, video, and audio. The model can generate up to 15 seconds of video at 2K resolution with native stereo sound. H3 excels at instruction following, accurate text and brand rendering, and V2V motion transfer. Early testing shows that it is ready for commercial content creation across a wide range of use cases.

GEMINI LIVE API OVERVIEW

The Gemini Live API enables low-latency, real-time voice and vision interactions with Gemini. It processes continuous streams of audio, images, and text to deliver immediate, human-like responses. The API creates a natural conversational experience for users. It can be used to build real-time agents for a variety of industries.

AGENT BEHAVIOR

Agent Behavior is an open standard for defining and evaluating how an AI agent should behave across a whole trajectory. Each behavior spec is a Markdown file that describes the recurring conduct that makes the agent reliable. The spec gives reviewers, rubrics, scorers, and evals something concrete to measure against. It can be used to review traces, write eval cases, revise prompts or tools, and communicate intended agent conduct to teams.

GPU MANAGEMENT: WHY IDLE GPUS ARE THE NEW GROUNDED AIRCRAFT

AI faces a bottleneck with GPU utilization, echoing how airlines maximize aircraft use for profitability. Owning more GPUs doesn't ensure efficiency - companies must focus on optimizing GPU workload orchestration to maximize output. Specialized models and continuous GPU management are crucial strategies for improving utilization and staying competitive in the AI arena.

WITH MOONSHOT'S FREE KIMI K3, CHINA CHANGES THE SOVEREIGN AI PLAYBOOK

China's Moonshot AI open-sourced Kimi K3 on July 27. Any government, company, or individual can now run the model on their own computers for free. They can also retrain the model however they want. The existence of a high-quality open model increases return on hardware investments because of a reduction or elimination of ongoing licensing costs.

ANTHROPIC SAYS ITS CLAUDE MODELS ‘GAINED UNAUTHORIZED ACCESS' TO OTHER ORGANIZATIONS' SYSTEMS

Anthropic discovered three instances where its models accessed the internet during an evaluation and gained unauthorized access to the systems of three different organizations.

HUGGING FACE STORAGE BUCKETS

Hugging Face Storage Buckets allow developers to store models, datasets, and artifacts with simple per-TB pricing.

JUDGE QUESTIONS ANTHROPIC SUPPLY-CHAIN RISK

A judge said the Trump administration had not provided enough evidence to classify Anthropic as a supply-chain risk or justify blocking federal agencies from using its technology.

TEACHING AN OPEN MODEL TO DO SCIENCE

Loka, Arcee, AWS, and Prime Intellect post-trained Trinity Mini with reinforcement learning across tool-assisted biomedical research and Gene Ontology annotation.

NVIDIA EXEMPLAR CLOUD: LESSONS FOR UNLOCKING FULL PERFORMANCE ON AI INFRASTRUCTURE

NVIDIA has identified configuration issues causing AI cluster performance gaps, such as CPU power settings and network tuning differences.

SUPERLOGICAL

Superlogical plans to build a composable multiplexer for all work that is safe and operable in production.

THINKING MACHINES COFOUNDER JOINED OPENAI

Thinking Machines co-founder Lilian Weng left the startup after citing health effects from sustained stress and workload, then joined OpenAI. She said the pace required by the startup had become physically unsustainable.

SPACEXAI LAUNCHES GROK VOICE THINK FAST 2.0 ON AGENT BUILDER

Grok Voice Think Fast 2.0 is now available at $0.09 per audio minute. grok-voice-latest will switch over to the new model on August 5. The release makes Grok Voice more dependable in real customer workflows.

HOW ENABLING TWO SETTINGS TRIPLED OUR SCORES ON THE ARC-AGI-3 BENCHMARK

GPT-5.6 Sol scores just 7.8% on the ARC-AGI-3 benchmark despite having solved longstanding open problems in mathematics and beaten games like Pokémon FireRed. Benchmarks rarely measure AI models in isolation. They also measure less visible choices about API settings, harness design, and prompting. Researchers discovered that turning on retained reasoning and compaction in ChatGPT and Codex tripled scores and cut output tokens by 6x on the benchmark.

DEEPMIND WON A NOBEL FOR ALPHAFOLD. THEN IT BROKE UP THE TEAM

The team that built AlphaFold, which won Google DeepMind a Nobel Prize, has been taken apart. Most of the original AlphaFold-paper authors were reassigned over the past year. Nearly a quarter of them have left the company, with a few moving to Isomorphic Labs, an Alphabet drug-discovery spinout, and the stars going to Anthropic. The reorganization marks a turn away from the deep-science bets that made DeepMind's name toward the Gemini-powered AI scientist race.