AMZN net income last quarter was 85% valuation jump in Anthropic, where it holds $180b equity. AWS AI revenue is ~45% Anthropic and 28% OpenAI per Barclays.
Key points
Net income composition: 85% from Anthropic valuation jump
AWS AI revenue split: ~45% Anthropic, 28% OpenAI per Barclays
Sources
OpenRouter provider router biggest improvement announced
OpenRouter shipped its biggest improvement to the provider router since founding, changing routing to reward the best quality inference for the actual cost of a request. The change keeps requests cheap while being fast and reliable.
Key points
Shipped biggest improvement to provider router since founding.
New router rewards best quality inference for actual request cost, keeping requests cheap, fast, and reliable.
Sources
Faraday launch – company-specific AI for manufacturers
Faraday launches company-specific AI agents for manufacturers that run work end to end, from back office to shop floor. The team invites manufacturers of physical products to meet them.
Key points
Launch: Faraday (YC F26) announced.
Agents: AI runs work end to end, from back office to shop floor.
Sources
- Original post
- faradaycompute.com - Linked in the post
micronuclear reactor design approved for safe nuclear operations
micronuclear reactor design was just approved for safe nuclear operations. The post congratulates the achievement.
Key points
Approval: micronuclear reactor design approved for safe nuclear operations
Acknowledgment: congratulations from @DougBernauer and @torishiv
Sources
- Original post
- radiantnuclear.com - Linked in the post
SF Compute: compute market forces AI labs deposits, $10B-$20B
SF Compute says the current compute market forces AI labs to put down huge deposits upfront, contributing to $10B-$20B valuations.
Key points
Compute market pushes AI labs to put down huge deposits upfront
Deposits help spark sky-high valuations of $10B, $20B
Sources
Claude self-assessment flags misuse since June
Claude introduced a self-assessment technique that gives a preliminary safety assessment after each response. Since its deployment in June, it has surfaced misuse that other systems missed.
Key points
Self-assessment: Claude flags potential misuse in conversations with a preliminary safety assessment.
Deployment date: Self-assessment has been used since June.