Good morning. It's Tuesday, July 28, and we're covering NVIDIA's next-gen AI servers, how AI is fueling America's startup boom, and SSI's major compute partnership with NVIDIA.

Plus: Zeitgeist Poll on your favorite coding agent.

NVIDIA backs Safe Superintelligence with an investment after gaining rare access to the startup's closely guarded research. The long-term partnership gives Ilya Sutskever's company access to NVIDIA's next-generation Vera Rubin systems while expanding its compute capacity by an order of magnitude. The companies also plan to co-develop future AI hardware platforms.

Sam Altman says his biggest AI concern is not the technology itself but power becoming concentrated among a few companies or governments. His comments follow debate over Moonshot AI's open-source Kimi K3 model and growing calls to restrict Chinese AI. Microsoft, Meta, NVIDIA, Google, and OpenAI support open-source AI, while Anthropic declines to back them.

Microsoft unveils MAI-Cyber-1-Flash, a security model that delivers top-tier vulnerability detection at half the cost of its previous AI system. Integrated with the MDASH platform, it achieves a 96% CyberGym score by pairing the new model with GPT-5.4 for the most complex tasks. The system draws on more than 100 trillion daily security signals and insights from 1.6 million customers to continuously improve cyber defenses.

SecondBrain Note is Genspark's first hardware product: a credit-card-sized AI recorder that snaps onto your phone and turns meetings, calls, and ideas into organized memory. Paired with SecondBrain and Super Agent, it can retrieve past context, draft follow-ups, and help you act on what you captured. The limited first release is live now with 10% off.

Shop the limited first release and get 10% off today!

ChatGPT’s Codex automated returns and refund negotiations with customer support, securing refunds while saving hours of effort.

The coding agent field split wide open this year. Terminal agents, IDE forks, and autonomous cloud builders are all fighting for the same seat. Which one earned yours?

Early testing suggests NVIDIA’s upcoming Vera Rubin AI server racks will be significantly easier for cloud providers to deploy than the company’s Grace Blackwell systems, which caused widespread installation challenges in 2025. Executives at several cloud providers say the new NVL72 racks benefit from a familiar third-generation design, though adapting software for large-scale deployments will still take several months.

The systems pack 72 GPUs and 36 CPUs, cost at least twice as much as the first Blackwell racks, and NVIDIA says they deliver up to 10 times greater efficiency in AI tokens per second per watt. Larger Vera Rubin Ultra systems arriving next year—with up to 576 GPUs and substantially higher power requirements—are expected to present a tougher engineering test, and some customers remain skeptical they will avoid the deployment issues that plagued Blackwell. → Read the full article here. (Paywall)

The United States is experiencing a sustained surge in new business creation, with Americans filing an average of 466,000 business applications per month since early 2024 and 531,000 in June 2026—nearly double the monthly average in 2019. Economists attribute the trend to a combination of a tighter job market, lasting shifts in consumer demand, and AI tools that make it easier to launch and run a business.

Surveys show 60% of founders used AI to help start their businesses in 2025, up from 2023, while e-commerce platforms continue to expand opportunities for niche retailers. Although many of these new ventures are one-person operations, some entrepreneurs are already scaling and hiring, suggesting AI may lower the barriers to starting a business without limiting long-term growth. → Read the full article here. (Paywall)

OpenAI is in talks with NVIDIA on a financing backstop of up to $250 billion to support debt funding for a planned 10-gigawatt AI data center campus in Pike County, Ohio, according to people familiar with the discussions. The proposed guarantee would allow OpenAI to raise lease and construction financing using NVIDIA’s credit, while separate negotiations continue over supplying the chips for the facility.

The campus, being developed with SoftBank, SB Energy, and the U.S. Department of Energy, could ultimately cost more than $500 billion and rank among the world's largest AI infrastructure projects. The discussions remain ongoing and could change, underscoring the scale of investment required as leading AI companies race to secure computing capacity for next-generation models. → Continue reading here.

$71M for Robot AI: Enigma lets users control 100 AI robots online to learn how people naturally interact with machines.

Google AI Search Becomes the Default: AI Overviews now appear in 43% of searches.

AI Data Center Sues Over Water Access: California project seeks 287M gallons a year from the Colorado River.

AI 'Doctors' Spread Misinformation: AI-generated TikTok accounts are pushing false health advice.

Thanks for reading. See you next time!

— Matthew Berman, Nick Wentz & the Forward Future Team