← Back to archive

Sunday, May 31, 2026

Anthropic now worth more than OpenAI

Anthropic just leapfrogged OpenAI with a staggering $965B valuation after raising $65B in Series H funding (wild), while researchers showed they can strip safety guardrails from Meta and Google models in literal minutes (yikes). Meanwhile, OpenAI's building tax agents that learn from corrections and Sakana AI dropped DiffusionBlocks to slash training memory costs. Would you trust a self-improving AI with your taxes?

Top Stories

1
Anthropic Is Now More Valuable Than OpenAI

Anthropic raised $65 billion at a $965 billion valuation, making it more valuable than OpenAI, with revenue reaching a $47 billion run rate driven by its Claude Code assistant. The funding comes amid preparations for IPOs across major AI labs including OpenAI and SpaceX/xAI.

anthropicopenaifundingvaluation
2
Self-improving tax agents with Codex

OpenAI

OpenAI and Thrive Holdings built Tax AI using Codex to create self-improving tax preparation agents that automatically learn from reviewer corrections, trace errors, and test improvements before deployment.

openaicodexagentsenterprise-ai-adoption
3
Anthropic Raised $65B in Series H Funding

Anthropic

Anthropic raised $65B at a $965B valuation with $47B run-rate revenue, underscoring massive enterprise demand for AI and intensifying competition among foundation model providers.

anthropicfundingllmenterprise-ai
4
AI guardrails stripped from Meta and Google models in minutes

Financial Times

Publicly available tools are being used to strip safety features from open-source AI models like Meta's Llama and Google's Gemma in minutes, with modified versions downloaded 13 million times. This undermines attempts to regulate AI at the point of development as open-source models approach frontier capabilities.

open-sourceai-safetymetagoogle
5
DiffusionBlocks

Sakana AI

Sakana AI's DiffusionBlocks enables training deep neural networks block-by-block rather than end-to-end, drastically reducing memory requirements while matching traditional performance across multiple model architectures. This breakthrough could help overcome resource limitations in AI training.

diffusion-modelstraining-efficiencymemory-optimizationsakana-ai

Keep Reading

Industry Voices

Enjoyed this issue?

Get daily AI intel delivered to your inbox. No fluff, just the stories that matter.