OpenAI clarified that its $70 billion annualized revenue figure reflects net revenue rather than its actual run rate near $50 billion, highlighting accounting discrepancies across major AI firms ahead of potential public offerings. Meanwhile, global AI dynamics continue to shift as Meta restricts ByteDance advertising, Chinese tech giants expand model investments, and US policymakers establish a new leadership team to oversee national AI strategy.
Nvidia is reportedly in negotiations to purchase or expand its ownership stake in US-based startup Reflection AI following an initial $800 million investment. The potential deal could provide the startup with massive hardware resources and cloud compute credits to advance frontier model development.
SpaceX has agreed to acquire a nationwide portfolio of up to 14 MHz of paired spectrum in the 800 MHz band from Grain Management to help Starlink become a major US mobile carrier.
A legal partner for server firm Super Micro pleaded guilty to facilitating the illegal diversion of restricted Nvidia AI hardware to China, highlighting enforcement actions around export controls. This case underscores systemic challenges in preventing advanced American processors from reaching foreign adversaries.
Nvidia has allegedly halted production of the GeForce RTX 5090 to prioritize AI data center and professional GPUs, creating an impending supply drought for gamers.
The Chinese Communist Party and State Council have directed officials to build a state-managed distributed ledger system to integrate national economic sectors. This centralized infrastructure aims to synchronize data management and transaction transparency across government and private enterprises.
Postman partnered with AWS to discuss the architectural patterns behind Agent Mode, an AI-native interface built using Amazon Bedrock to support 40 million developers. The post details strategies for managing tool sprawl, schema-based data reads, and context windows at scale.
The Allen Institute for AI (Ai2) rolled out a new GPU cluster scheduler incorporating time budgets, hierarchical fair-share allocation, and a time-slicing contract to optimize large-scale distributed training workloads and reduce human maintenance toil.
An open-source, fully local voice agent built in Rust demonstrates the inner workings of speech recognition, vector embeddings, and language model tool calling.
Research into using autoregressive diffusion models combined with flow matching and atom smoothing to synthesize realistic financial market data order book events.
Goodfire researcher Eric Bigelow joins the show to investigate how large language models arrive at decisions at the level of mechanistic interpretability. Drawing on his research into forking paths,…
A lot of people in AI treat evolution as a dumb fallback, basically random search for when you can't take a gradient. Akarsh Kumar thinks that is wrong. Selection hangs on to partial solutions, so…
From the Bitter Lesson of AI scaling to the unsolved mysteries of protein folding, Google DeepMind’s Pushmeet Kohli and Biohub’s Sal Candido are rethinking what it takes to build AI that truly…
Is the future of frontier AI open or closed? ReflectionAI co-founder and CEO Misha Laskin joins Sarah Guo and Elad Gil to discuss the launch of Beam, the company’s 500 billion parameter open-weight…
Nathan Labenz reports back from The Curve conference, sharing off-the-record insights from frontier lab leaders on shortened timelines and whether humanity is approaching an intelligence threshold it…
OutSystems CEO Woodson Martin joins Nathan to discuss how enterprise software can evolve at AI speed without breaking critical operations. Martin explains how an intermediate layer of abstract…
In this episode, Diogo Almeida, co-founder and CEO of TypeSafe, joins us to discuss Jev, TypeSafe’s recently released model for bringing fast, reliable intelligence directly into software. We explore…