AI Sector Daily Digest — June 6, 2026

President Trump signed National Security Presidential Memorandum-11 (NSPM-11) on June 5, replacing the Biden-era NSM-25. The directive orders the Defense Department, intelligence community, and 13 other federal departments to accelerate AI adoption across warfighting and intelligence operations. The core mandate: agencies must onboard the most advanced commercial AI models from multiple vendors, with a classified annex due in 90 days. The order also creates hard limits — AI must never be used to restrict speech, embed ideological biases, or conduct unauthorized surveillance of US citizens, with commanders held personally liable. Procurement processes get a 120-day deadline to enable rapid multi-vendor AI access.

Supabase closed a $500M Series F on June 5, led by GIC, with Stripe, Georgian, and Salesforce Ventures joining. The raise doubled its valuation from $5B to $10.5B in eight months. The catalyst: AI coding assistants have driven database launches on Supabase up over 600% in the past year, with over 60% initiated by AI tools. The platform now claims nearly 10 million developers as users, and CEO Paul Copplestone credits Claude Code and Codex specifically for expanding who can ship software. Also this week: Supabase launched Multigres, an orchestration layer that makes managing Postgres at scale significantly less manual.

On June 5, OpenAI announced six new enterprise plugins for Codex covering sales, data analytics, creative production, product design, public equity investing, and investment banking. Two additional features shipped alongside: Sites, which lets teams build and share internal mini-apps with live external data connections, and Annotations, which enables fine-grained edits on generated outputs. Non-developer users now make up about 20% of Codex's user base and are growing at more than 3× the rate of developers — the expansion signals OpenAI is building Codex as a general enterprise automation layer rather than a developer-only coding tool.

New York-based Flourish raised $500M at a $2.5B valuation in a round anchored by Jeff Bezos at approximately $100M, with Lux Capital and Google Ventures also participating. The startup is building what it calls "Cortex AI" — an architecture inspired by the continuous learning mechanisms of the human brain — with the stated goal of matching GPU-based AI performance at a fraction of the energy cost. Flourish is one of the few foundational AI bets targeting the power-consumption problem directly at the hardware-architecture level, rather than through model compression or inference optimization.

The Computer Vision and Pattern Recognition conference opened June 5 in Denver with 16,092 submissions — a 24% increase over 2025 and the highest in the event's history. Accepted papers: 4,089, up 42% from 2,878 last year, with a 25.4% acceptance rate. The most striking shift: vision-language and multimodal large language model papers now account for 10.6% of highlighted work, up from 4.9% last year. Taken together, generative and multimodal research makes up 22% of highlighted papers this year versus 14% last year, suggesting the field's center of gravity has moved decisively toward models that reason across modalities.