TLDR AI 2025-10-29

In Metronome's upcoming webinar, Martin Casado, General Partner at Andreessen Horowitz, reveals how leading startups and public companies alike are building monetization operating models for the AI era. You'll learn:

⏩ Why AI is accelerating the shift from traditional subscriptions to results-based and hybrid monetization models

💵 How fast-growing AI companies are building their Monetization Operating Models to align product, finance, GTM, and engineering

🤔 Why treating pricing like a product enables experimentation and growth

Save your seat

OpenAI has restructured into a non-profit with a controlling stake in its for-profit business. The nonprofit is called the OpenAI Foundation. It holds a 26% stake in the for-profit arm. The for-profit arm is now a public benefit corporation called OpenAI Group PBC. Microsoft holds an investment in the PBC that is roughly 27% of the company. Microsoft can now pursue AGI independently or in collaboration with third parties, and OpenAI can now jointly develop some products with third parties.

PayPal's digital wallet will be embedded into ChatGPT so users can pay for items through the chatbot starting next year. The deal will also allow merchants to sell items on the platform. The move is an early part of OpenAI's efforts to broaden ChatGPT's use for e-commerce. The company believes its 700 million-plus weekly users will use the service to find items, similar to a human personal shopper. PayPal is attempting to position itself as a payments backbone for the upcoming era of agentic AI shopping.

India is OpenAI's second-largest market with 29 million downloads in 90 days, but accounts for only $3.6 million in revenue. Perplexity and Google have also deployed aggressive free-tier strategies to capture the country's 700 million smartphone users to generate advertising revenue.

Substrate, a San Francisco-based startup founded by a protege of investor Peter Thiel, claims it has reinvented a key link in the microchip production process. It plans to use its innovation to build manufacturing facilities in the US. The startup has targeted a three-year timeframe for producing its first chips. Industry analysts say the high costs and enormous technical challenges in lithography give ASML, the sole global manufacturer of advanced lithography machines, a 10-year head start over any would-be competitors.

Meta's capital expenditure ratio hit 37% of revenue in Q3 2025, the highest capex-to-revenue ratio in the company's history. It announced $75.5 billion in infrastructure deals between September and October. Meta is financing the infrastructure for superintelligent AI systems. Meta can't compete with OpenAI on model quality alone - it needs a lot of dedicated compute, exclusive access to the latest Nvidia chips, and a vertically integrated stack to iterate faster, train bigger models, and serve more users. If AI doesn't deliver a return on Meta's investment, Meta will still have to pay rent for the next 16 years.

A forthcoming Nvidia chip would level the playing field between Chinese and US AI labs, which Trump may use as leverage over Xi in wide-ranging trade talks. The move would shrink America's massive AI compute advantage dramatically, especially since China's domestic alternatives are still far behind the US. Export restrictions are the only way to maintain the compute gap.

Last week, Baseten announced 650 TPS and 0.11 TTFT on gpt-oss, the most performant API on Nvidia hardware. Get near limitless scale and 99.99% uptime on one of the best open source alternatives to GPT-5.

• ⚡ The fastest TTFT available
• 🧪 650 TPS with nearly unlimited scale

Test gpt-oss today

FARMER integrates normalizing flows and autoregressive models to estimate likelihoods and synthesize images directly from pixels. It uses a novel dimension reduction and distillation strategy to improve efficiency while maintaining fidelity.

Agent HQ transforms GitHub into an open ecosystem that unites every agent on a single platform. Coding agents from Anthropic, OpenAI, Google, Cognition, xAI, and other providers will become available directly within GitHub as part of the GitHub Copilot subscription over the coming months. Agent HQ's power comes from a unified command center called 'mission control' that allows users to choose from a fleet of agents, assign them work in parallel, and track their progress from any device, wherever they work. It also provides new branch controls, identity features, one-click merge conflict resolution, and new integrations for Slack and Linear.

Eli Lilly and Nvidia have teamed up to build a supercomputer to boost the discovery of new medicines. The supercomputer and its AI capabilities will be able to identify new molecules to speed up the typical years-long development timelines. It will also help improve clinical trials, manufacturing processes, and sales. Financial terms of the partnership have not been disclosed. The system is expected to be online by January.

OpenAI is tracking toward achieving an intern-level research assistant by September 2026 and a fully automated 'legitimate AI researcher' by 2028. The AI researcher will be a system capable of autonomously delivering on larger research projects. OpenAI is betting on continued algorithmic innovation and dramatically scaling test time compute to achieve its goal of creating superintelligence within the next decade. The company's recent restructure will help support its aggressive timeline for AI research assistants while maintaining a commitment to responsible AI development.

Perplexity has outlined its practical vision for workplace AI, emphasizing usefulness over hype. With features like Comet, email and voice assistants, and app connectors, the platform helps reduce distractions, scale user impact, and streamline work toward meaningful outcomes.

Ragie is a fully managed RAG pipeline with multimodal support, structured chunking, and multi-index retrieval, delivering the right context to your LLM every time. Try Ragie for free

Shield AI will unveil its AI-piloted X-Bat fighter drone, which features vertical takeoff, a 2,000-mile range, and AI piloting via its Hivemind software.

Arvid Lunnemark worked at Stripe before co-founding Cursor in 2022.

Google, through Broadcom, has raised its planned monthly TSMC CoWoS wafer starts for next year.

The 2.9% investment makes Nvidia Nokia's second-largest shareholder.