Monday, May 4, 2026
OpenAI drops GPT-5.5 (and revenue targets)
OpenAI just dropped GPT-5.5 'Spud' with 35x cost reduction and autonomous multi-step capabilities, while Anthropic's Claude agents are literally negotiating and closing real deals now (wild). Meanwhile, OpenAI missed revenue targets and dragged Oracle and chip stocks down with it, even as the FDA announces they're using AI to slash drug trial times. Here's the kicker though: when Claude agents negotiated those deals, better models consistently won—and humans on the other side had no idea they were at a disadvantage. Would you want to know if you're negotiating against an AI?
Top Stories
Bloomberg
The FDA plans to cut drug development times by months or years through real-time monitoring of clinical trials using AI, tracking efficacy and safety signals without full patient data collection.
OpenAI launched GPT-5.5 ('Spud'), a significantly more efficient model that reduces costs by 35x per token and handles complex multi-step projects autonomously, positioning it as a direct competitive response to Anthropic while advancing their super app strategy.
Anthropic's experiment with Claude agents autonomously negotiating real transactions showed stronger AI models secured measurably better deals, but participants couldn't perceive the quality gap — raising critical questions about disclosure requirements and fairness as AI-mediated commerce approaches mainstream adoption.
CNBC
OpenAI's reported revenue and growth shortfalls sparked a broad selloff in AI infrastructure stocks including Oracle, chipmakers, and SoftBank, though analysts debate whether this reflects OpenAI-specific market share losses to competitors like Anthropic and Gemini or signals broader concerns about AI spending sustainability.
Anthropic
Anthropic's experiment with AI agents conducting real marketplace negotiations found that more capable models achieved significantly better deals, but users with weaker models didn't notice their disadvantage—highlighting potential risks of hidden inequality in AI-mediated commerce.
Keep Reading
Industry Voices
Keir Bradwell
Researcher at Anthropic
Shares technical insights on AI safety research and interpretability work happening inside Anthropic's labs.
Kevin K. Troy
Researcher at Anthropic
Posts about alignment research and the practical challenges of making AI systems behave reliably at scale.
Enjoyed this issue?
Get daily AI intel delivered to your inbox. No fluff, just the stories that matter.