Labs & Research
OpenAI, Anthropic, Google DeepMind, Meta AI and publications.
36 published articles
The deprecation dossier
Mistral AI just killed half its model family. Here is what survived and why.
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 by mid-2026. The company's strategy emphasizes small specialist models for agents and coding, while deprecating experimental variants like Magistrate and earlier Devstral versions.
2026-07-13
Artificial Intelligence
Meta bought Manus for its infrastructure, not its chatbot
Manus joins Meta in a strategic acquisition that prioritizes agent infrastructure over consumer AI. With over 80 million virtual computers created and 147 trillion tokens processed, Manus becomes Meta's execution layer for enterprise automation.
2026-07-13
AI Agents
Mistral coding agents leave your laptop behind: parallel sessions, no hovering required
Mistral is moving coding agents to the cloud, enabling parallel, asynchronous task execution via the new Mistral Medium 3.5 model. The update introduces remote agents in Mistral Vibe and Le Chat, along with a Work mode for multi-step tasks, while keeping human oversight for sensitive actions.
2026-07-12
Acquisition
Mistral buys its way into physics, and a race against NVIDIA and Ansys
Mistral buys its way into physics simulation, acquiring emmi AI's neural surrogates for CFD and plasma turbulence. The deal puts Mistral in direct competition with NVIDIA and Ansys for the industrial engineering market.
2026-07-11
Internal research update
OpenAI's AGI roadmap leans hard on voice and vision. The safety part is harder.
OpenAI's updated research roadmap runs from GPT-5 through multimodal generation to alignment work. The ambitions are huge; the safety section still reads like an afterthought.
2026-07-11
Multimodal AI
ViQ just gave multimodal AI the one thing it needed: discrete tokens that don't lose detail
ViQ tackles a core trade-off in multimodal AI: discrete visual representations either lose semantic meaning or sacrifice detail. Tencent-Hunyuan's two-stage approach delivers competitive performance with continuous encoders while slashing training time by up to 70 percent.
2026-07-11
AI Research
DiScoFormer found a way to kill the AI bottleneck that KDE and neural nets both missed
DiScoFormer estimates density and score in one forward pass, beating KDE by 37x in density error at 100 dimensions. The same transformer could serve diffusion models, Bayesian inference, and particle simulations without retraining.
2026-07-11
Grok 4.5
Cursor's Grok 4.5 was built by AI agents, not humans. That's the real story.
Cursor's Grok 4.5 is a Mixture-of-Experts model built using reinforcement learning in environments created by earlier AI agents, not humans. It handles complex, long-duration tasks across software engineering, data science, finance, and law, and it's available now.
2026-07-10
AI Model Launch
Anthropic Launches Claude Sonnet 5: A More Agentic Model at Competitive Pricing
Anthropic releases Claude Sonnet 5, a model that closes the gap with Opus-class models on agentic tasks like coding, tool use, and reasoning. Priced competitively, it launches today across plans and the Claude API.
2026-07-10
Event Announcement
Anthropic Launches Claude Founder House in Paris to Unite Europe's AI Startup Ecosystem
Anthropic launches the Claude Founder House in Paris, a one-day event uniting Europe's top AI startup founders and builders for keynotes, fireside chats, and hands-on development sessions with Claude.
2026-07-10
Special Report
Meta AI's open-source bet just broke the business model of its rivals
Mark Zuckerberg is spending billions on a bet: that the best way to build AI is to give it away. This is the story of the lab leading the charge, the people, the philosophy, and the debate that won't fade.
2026-07-10
Portfolio strategy
Alibaba's Qwen is building a model for every AI job, not just one to rule them all
Alibaba Cloud's Qwen team quietly released a 35B MoE agent world model, three new ASR models, and an image-generation RL report, revealing a strategic bet on breadth over spectacle in the AI model race.
2026-07-09