Possible

Reid Hoffman sees promise in an AI buildout and wants its financial risks in the open. He and Aria Finger unpack a Financial Times report on up to $300 billion in AI infrastructure guarantees largely absent from the balance sheets of technology companies. They explore California Governor Gavin Newsom’s push for independent lab auditors and an AI kill switch, and discuss what effective federal oversight could look like. Reid explains how these proposals differ from the 2024 bill he opposed and makes the case for an international cyber treaty. A Truth Social poll to rename AI also prompts a debate about why “artificial” sells the technology short.

Can America lead in AI while making life better for everyday Americans? Reid Hoffman and Aria Finger weigh in on Dario Amodei’s call to pace the frontier, and Reid makes the case for America’s dual mandate: leading in AI while delivering prosperity for the many. Anthropic researcher Evan Hubinger puts his personal estimate of AI-driven extinction risk above 10%. They unpack that warning alongside Sam Altman’s support for slower development and David Sacks calling Anthropic and OpenAI a cartel. Reid identifies biological threats, cyberattacks, and job disruption as the real risks, makes the case for steering over pausing, and explains why kill switches matter. From data centers that don’t raise local electricity bills to a free AI medical assistant for every American, they explore what rebuilding public trust could require.

Read “AI for All Americans” on Substack: https://reidhoffman.substack.com/p/ai-dual-mandate

Reid Hoffman and Parth Patil argue that if your AI experiments always work, you’re not pushing hard enough. Wrapping up the Tokens to the Future series, they revisit seven grantees’ experiments with $1,000 a week each in tokens. Their conversations span Dungeons & Dragons worldbuilding, an iOS game built in a weekend, and personal agents whose usefulness compounds over years. They explore how AI can shift more people from execution-focused “cog jobs” to creative “spark jobs,” and why expertise can sometimes make new possibilities harder to see. Throughout, they ask what to delegate and where story, pacing, and judgment still demand human direction. Plus, Parth’s example of ChatGPT picking seven for a “random” number raises a bigger question: how do you get surprise from a system built to predict?

When AI can fake your CFO’s face and voice, what still counts as proof of identity? Infosec Analyst Dozie Anazia joins Reid Hoffman and Parth Patil to explain how AI is transforming cyberattacks, penetration testing, and autonomous defense. Drawing on projects including the Breakfast news app and Aegis URL scanner, he shows how Claude and Codex can build software, uncover an XSS flaw, and run a pen test—while still falling short of unsupervised network defense and elite human hackers. They explore purple teaming, agent-versus-agent review, and the advantage defenders retain over attackers. Plus, one low-tech defense against a stolen voice: a family safe word.

Useful robots won’t be programmed one task at a time; they’ll need to adapt to unfamiliar objects, environments, and robot bodies. Physical Intelligence cofounder Chelsea Finn (named last month to the TIME100 AI list for 2026) joins Reid Hoffman and Aria Finger to explain how general-purpose robot models learn to act in the messy physical world. She explores why lower-quality training data can strengthen a model, what months of failed laundry-folding attempts revealed, and how the same system can work across different robot platforms. A robot mistaking an oven for a drawer becomes a lesson in the strange line between useful generalization and obvious error. They also examine why robot demos can mislead, how hardware reliability can stall progress, and what it will take to move AI off the screen and into the world.

Congratulations to Chelsea, recently named one of TIME’s 100 most influential people in AI:

https://time.com/collection/time100-ai/2026/chelsea-finn/