David Senra: his podcast episode with Ben Wilson

Senra asked people to watch his new episode with Ben Wilson, host of How to Take Over the World. He says Wilson has stage 4 cancer and four kids under 5. The episode covers what Wilson has learned from studying history's greatest leaders and, from 55:51, how he is facing cancer @davidsenra. Senra called Wilson "one of my favorite podcasters" and asked viewers to help with "medicine, introductions, or donations" @davidsenra. Patrick O'Shaughnessy shared the appeal and said Ramp, Deel and AppLovin had each just given $100k toward life-extending medicine and support for Wilson's family @patrick_oshag.

O'Shaughnessy also posted a clip from his conversation with Nubank's David Vélez. In it, Vélez predicts the world's roughly 190 currencies will shrink to about 10 within 10–20 years, and calls US stablecoins "one of the biggest saviors of the US dollar" @patrick_oshag.

Two write-ups on agent sandboxes

• Guillermo Rauch shared Vercel's findings from its $1M Vercel Sandbox escape challenge. The challenge drew 1,285 reports, paid $242,000 in bounties and turned up one Linux kernel zero-day, with no reports of customer data being compromised @vercel. Rauch says the OpenAI agent escape and the Hugging Face hack made "agent containment" a top priority for AI infrastructure companies. He also says a white-hat researcher "armed with a swarm of AI agents" got through the Linux virtualization boundary that "nearly every cloud sandbox provider" depends on @rauchg.
• Aravind Srinivas pointed to his company Perplexity's write-up on its SPACE sandbox. SPACE runs millions of executions a day at a low error rate while spreading work across multiple machines to cut costs @AravSrinivas. The post explains how cryptographic identity keeps each request tied to the right execution as sandboxes pause, resume and move. It puts the error rate below 0.0002% @perplexity_ai.

Vinod Khosla: Muse, and his own 2017 blog

Khosla linked Exponential View's piece on Meta's Muse as a "digital butler." He argues that agents will either work for a corporation and treat the consumer as the product, "as Muse will be," or work for the consumer, "as in Fo" @vkhosla. He also reposted his 2017 blog post, "AI scary for the right reasons." His reason for resurfacing it: China dismisses AI dangers, and Putin has said whoever wins AI rules the world. He calls these geopolitical risks the biggest dangers, while saying "AI going wild is also a real danger" @vkhosla.

Shorter mentions

• Scott Belsky shared Sahil Bloom's "30 Hours in a Massive Earthquake" as "some perspective" from the Panama earthquake. He framed it with Cormac McCarthy's line: "You never know what worse luck your bad luck has saved you from" @scottbelsky.
• Martin Casado replied "Same." to @poteto's post @martin_casado. Poteto says he never lets agents write for him because "writing is thinking," and that knowing the right words also makes it easier to direct agents @poteto.
• Chamath Palihapitiya shared 8090 Factory's video with the line "Learn the past to rewrite the future…" @chamath. The video describes mapping what a legacy system can do, drawing on its old code, vendor systems and long-time operators, before AI agents touch it. People and agents then work from that map through a governed gateway such as MCP @8090_Factory.
• Bill Gates shared a Times article on Andy Burnham and a British dementia cure. His comment: "This kind of commitment will help turn promising research into breakthroughs" @BillGates.
• Reid Hoffman, in a video conversation, praised Gates's letter for pointing out that people "aren't working enough on economic transformation." He calls for "break glass" unemployment-insurance plans and for using AI to help people find and learn new work Mo News.
• Alexandr Wang called Claire Vo's How I AI review of Muse Wonder and Wonder Full "thoughtful." Wang is promoting his own lab's model @alexandr_wang. Vo "loved how it handled agentic tasks, really hated what it designed" (full episode) @clairevo.
• Morgan Housel shared his and Craig's podcast with Matt Sonsini on philanthropy, raising capital and how Silicon Valley has changed (Apple, Spotify) @morganhousel.
• Elon Musk replied 🎯 to François Chollet's post @elonmusk. Chollet argues that "intelligence is practically never the bottleneck"; the bottleneck is "iterated contact with reality" @fchollet.