AI
Artificial intelligence: LLMs, agents, diffusion, vision, NLP and the latest from top labs.
593 published articles
Generative AI
An AI that can see its own mistakes and undo them mid-generation
Google introduces CO2Jump, a training-free sampler for joint text and image generation. It uses a self-correcting Markov jump process where each modality's confidence scores guide the other's updates in real time, catching cross-modal errors mid-generation.
2026-07-14
Physical AI
Claude just went to work in a chip factory. The chat era is over.
Anthropic and UST are putting Claude to work inside semiconductor and manufacturing validation pipelines, training 20,000 engineers. The partnership shows how LLMs can catch design flaws earlier and compress four-day turnarounds into 48 hours, a shift that moves AI beyond text and into the production of physical goods.
2026-07-14
Google DeepMind
Gemma 4 just made every other open-weight model look 10x too big
Google DeepMind's Gemma 4 natively multimodal open-weight family introduces thinking mode, encoder-free architecture, and MoE options. The 2.3B model matches Gemma 3's 27B performance. The 31B model tops open-weight leaderboards.
2026-07-13
OpenAI vs Anthropic
OpenAI just made workers you can schedule. Claude is still a chat.
OpenAI just made workers you can schedule: persistent, shared AI agents that run in the cloud and integrate with Slack. Anthropic's Claude, focused on single-session reasoning, has no answer yet.
2026-07-13
Portfolio strategy autopsy
Mistral killed half its model family. What survived tells you everything.
A mid-2026 audit of Mistral AI's model portfolio reveals 36 active models across frontier, specialist, and legacy tiers, with 19 models slated for deprecation. The company's strategy emphasizes small specialist models for agents and coding, while retiring experimental variants like Magistrate and earlier Devstral versions.
2026-07-13
Artificial Intelligence
Meta didn't buy Manus for its chatbot. It bought the pipe that delivers 147 trillion tokens.
Meta acquires Manus for its agent infrastructure, not its chatbot. With 80 million virtual computers created and 147 trillion tokens processed, Manus becomes Meta's execution layer for enterprise automation.
2026-07-13
Cloud Infrastructure
Alibaba is building both the brain and the immune system for AI agents
Alibaba Cloud unveils a codebase-aware retrieval system for AI coding and an agentic WAAP security platform, both leveraging large models to solve latency, privacy, and compliance gaps. The hybrid architecture aims to outpace embedding-only tools and legacy WAF vendors.
2026-07-13
Benchmark integrity
GPT-5.6 Sol almost cracked a physics benchmark built so AI couldn't cheat
GPT-5.6 Sol (max) leads CritPt, a new physics benchmark built from unpublished graduate-level research problems, scoring roughly 5 points above GPT-5.5 and 4 points ahead of Claude Fable 5, but even the winner solved only a third of the problems.
2026-07-13
Remote Sensing AI
Smarter tokens just cut satellite AI costs by 3x without losing accuracy
Ai4earth's OlmoEarth v1.1 cuts compute costs by up to 3x over v1 for satellite image analysis, using a smarter token merging technique that maintains performance. The updated models enable cheaper planet-scale map refreshes for partner organizations.
2026-07-13
Cognitive debt in education
Homework scores are up 18%. Exam scores are down 20%. The gap is cognitive debt.
A 26,000-student study, an MIT brain-imaging experiment, and new OECD research converge on the same finding: generative AI raises homework scores while quietly eroding exam performance and neural engagement, unless teaching is redesigned around it.
2026-07-13
Spreadsheet agent review
Kimi sheets writes the formulas. Most ai tools just hand you a recipe.
Kimi Sheets is an AI agent that turns plain-language instructions into finished spreadsheets, with real formulas, pivot tables, charts, and smart formatting. Unlike chatbots that only give advice, Kimi Sheets directly edits the file.
2026-07-12
Editor's guide
The ten rules that separate AI coding agents from expensive guesswork
A practical guide to working effectively with AI coding agents like GitHub Copilot, Cursor, and Codex-based tools: scoping tasks, reviewing diffs, matching tools to tasks, testing discipline, and keeping humans in the loop on shared code.
2026-07-12