Newsletter Archive

Nvidia is projected to capture roughly two-thirds of the x86 server CPU market by FY2027 with its Vera processors, generating $20 billion in revenue and upending Intel and AMD's dominance. The shift marks Nvidia's expansion from AI accelerators into the foundational compute layer.

Tom's Hardware · 1 min read Hardware

A 3-billion-parameter specialized model for structured OCR outperformed every commercial frontier API tested while costing 50 times less, challenging the enterprise AI assumption that larger models are always the safer choice. Dharma-AI's analysis shows specialization and distributional alignment matter more than raw parameter count in production systems.

Hugging Face · 13 min read Research

An OpenAI internal model has autonomously disproven a central conjecture in discrete geometry (the Erdős problem from 1946), with the proof verified by prominent mathematicians who call it a genuine breakthrough. The verifying mathematicians noted no previous AI proof has come close to this quality; one said they'd recommend acceptance to Annals of Mathematics without hesitation.

MIRI · 3 min read Research

Trump shelved a draft executive order on AI cybersecurity and safety after pressure from tech allies, leaving key questions unresolved on government access to frontier AI models and AI safety mandates. The industry and White House are now reassessing regulatory direction.

Axios · 1 min read Policy

OpenAI is preparing to file confidential IPO paperwork with a potential $1 trillion valuation and September listing date, following SpaceX's record offering. The filing will test whether public markets will fund the staggering infrastructure costs and losses at the heart of the AI race—and set the tone for upcoming IPOs from Anthropic and other AI heavyweights.

Fortune · 6 min read Market

Anthropic is closing a $30B+ funding round at a $900B+ valuation as soon as next week, overtaking OpenAI as the world's most valuable AI startup.

Bloomberg Tech · 1 min read Market

Trump abruptly shelved a voluntary AI model vetting executive order after last-minute pressure from Elon Musk, Mark Zuckerberg, and David Sacks, citing concerns it would slow U.S. dominance over China. The order would have required companies to submit advanced models to national security agencies for 90-day pre-release testing—something OpenAI and Anthropic had already agreed to.

Fortune · 7 min read Policy

Researchers dissect Google's claim that AI agents built an OS for $916, revealing the "single prompt" was actually thousands of lines, custom scaffolding was purpose-built for the task, and key details about attempts and specificity are missing—exposing how vendor demos can obscure actual model capability vs. engineering effort.

AI Snake Oil · 5 min read Community