AI digest: Big money moves and technical breakthroughs
Anthropic goes public, MiniMax drops an impressive open model, and Nvidia makes a serious play for consumer AI hardware.
The AI industry is hitting its stride with major funding moves and some genuinely impressive technical releases.
Anthropic files for IPO as AI companies chase public markets
Anthropic has filed confidentially for an IPO with the SEC, valued at nearly $1 trillion after its latest funding round. This follows Alphabet’s plans to raise $80 billion for AI infrastructure, showing how much capital these companies need to stay competitive. The timing suggests Anthropic thinks the AI hype has enough staying power to impress public investors.
MiniMax M3 brings million-token context to open weights
Chinese company MiniMax released M3, an open-weight model with a 1M-token context window and native multimodality. The model uses “MiniMax Sparse Attention” architecture and supports computer use alongside coding tasks. This is significant because it’s the first open-weight model to combine top-tier coding performance with such a massive context window.
Nvidia launches RTX Spark to challenge Apple on AI PCs
Nvidia unveiled RTX Spark chips combining Blackwell GPUs with Arm-based Grace CPUs, claiming 1,000 TOPS performance for local AI agents. Microsoft, Dell, HP, and others are building PCs around it. If Nvidia can make AI agents actually useful on consumer devices, this could be the breakthrough that brings AI beyond chatbots into proper computing workflows.
Memory OS adds persistent memory to AI agents
Researchers released Memory OS, a six-layer open-source memory stack built on Hermes Agent. The system adds local persistent memory with gated retrieval and wiki functionality. This tackles one of the biggest problems with current AI agents - they forget everything between sessions, making them useless for ongoing tasks.