AI
Artificial intelligence: LLMs, agents, diffusion, vision, NLP and the latest from top labs.
77 published articles
Climate Technology
Aimip phase 1: A new benchmark to test ai climate models
The AIMIP Phase 1 project, involving groups from NVIDIA, Google Research, and others, provides an open dataset and evaluation framework for AI climate models. While these models accurately reproduce historical climate patterns, their ability to generalize to unseen conditions remains a key challenge.
2026-07-03
Artificial Intelligence
Microsoft’s GridSFM predicts power grid flow in milliseconds, targeting $20B in congestion savings
Microsoft's GridSFM foundation model predicts AC optimal power flow in milliseconds, enabling real-time grid scenario analysis. The open-source version covers grids up to 4,000 buses, with a premier tier for production-scale systems.
2026-07-03
Memory Management
mimalloc, Microsoft's tiny memory workhorse, is quietly powering AI at scale
Microsoft Research's mimalloc memory allocator, designed for high concurrency and large memory scales, is quietly powering AI and cloud services at scale. Its thread-local design with thousands of free lists minimizes contention and delivers both speed and memory efficiency.
2026-07-03
Export Controls
Anthropic Restores Access to Claude Fable 5 After US Export Controls Lifted
Anthropic restored Claude Fable 5 access globally after US export controls were lifted on June 30. The company introduced a new safety classifier to block a reported jailbreak and proposed a consensus framework for rating AI jailbreak severity with partners including Amazon, Microsoft, and Google.
2026-07-01
AI Models
Anthropic Launches Claude Fable 5: A Fifth-Generation Model for Extended Autonomous Work
Anthropic introduces Claude Fable 5, a fifth-generation model that can run agents for days, tackle ambitious coding projects, and handle complex enterprise workflows with minimal oversight. Pricing starts at $10 per million input tokens.
2026-07-01
Virtual Event
Anthropic Announces "The Briefing: AI for Science" Virtual Event for June 2026
Anthropic's virtual event on June 30, 2026 will detail how Claude is transforming scientific research, with participation from pharma executives, biotech leaders, and academic institutions, highlighting AI-native research environments and safety architectures.
2026-07-01
Benchmark Deep Dive
ARC-AGI-2: The Benchmark That Measures Fluid Intelligence in AI Systems
ARC-AGI-2 tests AI systems on fluid intelligence through visual grid puzzles that can't be solved by memorization. Top frontier models now score 75-85%, but the grand prize of $700,000 remains unclaimed. Here's a deep dive into the benchmark's design, scoring, and current leaderboard.
2026-07-01
AI Benchmarks
GPT-5.4 Leads the 2026 Math Benchmark Pack as Frontier Scores Saturate
GPT-5.4 edges out rivals with a sweep of the top competition-math rows, but GPT-5.2 Pro ties every cell and GPT-5.3 Codex offers nearly identical scores at lower cost. The real differentiator: how models perform on next-generation benchmarks like FrontierMath and HMMT Feb 2026.
2026-07-01
AI Research
Token-Level Analysis Reveals Hybrid LLMs Excel at Meaning-Bearing Words, Lag on Repetitions
Researchers at the Allen Institute for AI compared Olmo 3 and Olmo Hybrid token by token, finding hybrids excel on meaning-bearing words but not on repeated tokens. The work suggests filtered token losses as a richer evaluation method for language model architectures.
2026-07-01
API Update
Anthropic Boosts Claude API Rate Limits, Streamlines Tiers to Three
Anthropic increases Claude API rate limits for Sonnet and Haiku, simplifies usage tiers to three (Starter, Development, Scale), and ensures no organization sees lower limits. Changes are automatic.
2026-07-01
Microsoft Build 2025
Microsoft Discovery Now Generally Available for Agentic AI Workflows in Science and Engineering
Microsoft Discovery is now generally available, offering organizations a platform to build and govern agentic AI workflows for scientific and engineering research. The companion Microsoft Discovery app preview aims to democratize AI-driven discovery for domain experts.
2026-07-01
Benchmark
Oxford's GauntletBench puts AI agents through 100 real tasks. They failed 81% of them.
Oxford's GauntletBench puts AI agents through 100 challenging real-world tasks. Frontier systems cap out at 19.1% success, far below human performance. The benchmark targets overlooked capabilities in temporal perception, graphical understanding, and 3D reasoning across five professional applications.
2026-07-01