KM4BIKM4BI

Big companies are hiring again, Anthropic’s open-weight position, Zuckerberg backs AI for all

TBPN covers the hiring rebound around AI, Anthropic’s open-weight stance, and Zuckerberg’s argument for broadly accessible AI.

The strongest signal in this episode is the shift in the AI-and-jobs narrative. After months of hiring freezes and layoffs often framed around AI, several large companies are now saying they need more people to work alongside AI systems. The hosts frame this as a practical version of Jevons paradox: when a tool makes work more efficient, demand can rise rather than disappear.

Their view is not that AI has no labor impact. Instead, it changes where work sits. Some one-off freelance-style tasks become easier to do internally, while core work still depends on responsible people who can use AI effectively and integrate it into teams.

The episode also runs through infrastructure and market signals: Recursive Superintelligence signs a $410 million compute deal with AWS, Nvidia becomes a tenant in a huge Texas data center that will use Nvidia chips, and Apple prepares an upgrade/leasing program for iPhone, iPad, Mac, and Apple Watch. Each item points to the same broader pattern: AI is reshaping capital allocation, compute demand, distribution, and device economics.

The central policy discussion is about open-weight models. Dario Amodei clarifies that Anthropic is not calling for a blanket ban, but supports three ideas: continued chip restrictions toward China, action against industrial-scale model distillation, and mandatory safety testing for sufficiently capable models. The hosts focus on the hard parts: proving distillation, setting risk thresholds, and deciding what government can do beyond what labs can already enforce through contracts and hosting controls.

They also highlight the regulatory trade-off. Mandatory reviews can be reasonable for truly dangerous systems, but they can also become a bottleneck for smaller teams if incumbents have better access to Washington and faster approval paths.

Zuckerberg’s Wall Street Journal piece rounds out the debate by arguing for broadly accessible AI. His case is that openness can improve safety over time, much as open source has often done in software. The real question is therefore not simply open versus closed, but how to handle genuinely dangerous models without freezing competition and innovation.

Source