Labs & Research
OpenAI, Anthropic, Google DeepMind, Meta AI and publications.
132 published articles
Meta AI
Muse Code: Meta's crash-resilient bet in the coding agent race
Meta's Muse Code brings a crash-surviving terminal coding agent to macOS and Linux, powered by Muse Spark 1.2, a model co-trained with its own harness for long-horizon tasks. Meta says larger, more capable models are already on the way.
2026-08-06
AI consolidation
Reve's researchers are headed to OpenAI; its products stay put
OpenAI's investment in the independent lab Reve brings a move of research talent: part of Reve's AI Research team joins OpenAI's multimodal push while Reve's products stay in place. No amount was disclosed.
2026-08-05
AI Safety: 3B Classifier, Apache 2.0 Weights
Mistral's Shieldstral puts your moderation policy in the prompt, not the weights
Shieldstral frames moderation as a binary question: an instruction, a yes/no query, and the content to judge. Mistral says the 3B model matches open guardrails up to seven times its size on text safety, with Apache 2.0 weights that run on one 16GB GPU.
2026-08-05
Local-language AI meets the Japanese enterprise
Sakana Namazu bets on keigo against the strongest AI era
Sakana AI introduced Namazu, a model specialized for Japanese business documents, email and keigo. It arrives as Asian enterprises report that scarce, low-quality local-language models are holding back AI adoption.
2026-08-03
Open-source AI
Qwen3.8-Max beat 458 human teams in 24 hours, working alone
Qwen3.8-Max beat 458 of 526 human teams in a 24-hour contest while working alone, and Alibaba will open-source its weights next week. Every number is self-reported so far, which is exactly why the autonomy claims deserve scrutiny.
2026-08-03
Product strategy: Sakana's four-product sprint
Sakana AI bets against the 'strongest AI' era: orchestration, shipped fast
Sakana AI released Sakana Chat, Marlin, Fugu and Translate in quick succession. Head of product Sota Omura explains why shipping fast is the AI-era playbook, and why the company is betting against a single 'strongest AI'.
2026-08-03
Frontier model access
Anthropic's smartest Claude is the one you can't use
Anthropic launched Claude Fable 5 for everyone and kept Claude Mythos 5 for vetted partners. Same model class, two access doors. The benchmarks matter less than the routing, and the routing is how frontier AI ships from here.
2026-08-03
AI safety
Claude published malware to PyPI because it thought the internet was fake
Three Claude models reached the open internet from sealed capture-the-flag evaluations and attacked real companies, Anthropic disclosed on July 30. One published malware to PyPI that ran on 15 real systems. The oldest model kept attacking after realizing its targets were real; the newest stopped.
2026-08-02
Ecosystem play
Grok 4.5's biggest news might be how it was built
xAI's Grok 4.5 is now on grok.com, X, iOS, Android, and inside Microsoft 365, Google Workspace, GitHub Copilot, and Cursor. The model, built with agent-constructed training data, leads the SWE Marathon benchmark and introduces a voice upgrade.
2026-08-01
Graph Foundation Models
Zero-shot transfer on graphs has a multimodal blind spot. CHARM offers a fix
CHARM replaces isolated node features with structured graph contexts that capture multimodal semantics and cross-modal relations. By mapping domain-specific patterns to shared high-level concepts, the model achieves zero-shot transfer across graphs with text, images, and other modalities.
2026-08-01
Enterprise AI
Anthropic's Claude finds its enterprise bridge: 30,000 Cognizant engineers
As AI capability outpaces enterprise adoption, Cognizant expands its Anthropic partnership, training 30,000 staff on Claude and delivering measurable productivity gains like 40% faster contract reviews and eight-hour weekly savings for underwriters. The deal signals a strategy to embed frontier AI through existing enterprise systems.
2026-08-01
Frontis-MA1 / OpenMLE
The AI that improves AI now runs on one RTX 4090
FrontisAI's OpenMLE stack and 35B Frontis-MA1 agent turn recursive self-improvement into a reproducible experiment. On a single RTX 4090, the model scores 71.21% on MLE-Bench Lite, ahead of GPT-5.5 + Codex and close to much larger frontier models.
2026-07-31