AIToday
Hacker NewsPublished: May 23, 2026, 19:00 JST1 min read

Andrej Karpathy joins Anthropic as AI centralization debate intensifies; Pluralis releases Agora for distributed GPU training

Andrej Karpathy joins Anthropic as AI centralization debate intensifies; Pluralis releases Agora for distributed GPU training

3 Key Points

  1. Andrej Karpathy, co-founder of OpenAI and angel investor in Prime Intellect, announced he is joining Anthropic. Karpathy had written in 2023 about thinking "a lot about centralization and decentralization" but is now joining what the article describes as perhaps the most hostile major lab toward open-source AI, contrasting with OpenAI and Google, which have released open-weight models, while Anthropic limits access to its frontier model, Mythos.

  2. Pluralis Research released Agora, a pretraining stack that allows non-collocated consumer-grade GPUs to be competitive with centralized clusters. In a pilot run, Pluralis trained an 8B model using Agora; the technology is 15x faster than Megatron-LM in this setting and is only 1.5x less efficient in terms of tokens per unit compute. Consumer-grade GPUs must contain 24GB of VRAM and be located in North America.

  3. Nebius announced GPU rental price increases effective June 1: H100 rentals will cost $2.15/hr to $3.85/hr, up from $1.25–$2.95/hr, representing increases of 26% to 72% depending on conditions.

Ask the AI about this article →

Get AI news like this every morning

For example, today's edition would include:

  • Pasqal, fresh off Nasdaq listing, courts Taiwan photonicsDIGITIMES Asia · 1h ago
  • Lam Research exec: Automated preventive maintenance key to future fabsDIGITIMES Asia · 1h ago
  • Dell's bullish forecast eases AI demand fearsYahoo Finance AI · 1h ago

AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.

Free · takes 30 seconds · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. Q&As are published on this page for other readers too.

Next articleBofA research finds AI currently lifting macroeconomic productivity by approximately 0.1% per year, with gains visible in narrow corporate tasks but not yet scaling economy-wide