Mark Zuckerberg pushing back on AI slowdowns is a deeper illusion
By Ray Dombrowski · Reporting from Youngstown ·
TBPN argues that Mark Zuckerberg's vision of self-regulating AI labs is a deliberate sidestepping of existential risk.
On the latest episode of the podcast TBPN, hosts John Coogan and Jordi Hays tackled the twin shocks of the Federal Reserve’s first interest rate hike in three years and the escalating rhetorical war in Silicon Valley over artificial intelligence regulation. On the podcast, Coogan argued that the fierce backlash against third-party AI evaluators like METR misses a fundamental structural distinction, noting that the people writing frontier forecasts do not need to be the same people executing day-to-day safety mandates. Comparing the enterprise to the armed forces, Coogan noted: you can be 22 years old, not graduate from college, go into the Navy, enlist in the Navy, not even in the officer training program, and in 18 months you can be responsible for the security of a nuclear power plant on a submar. You do six months in nuclear field A school, about six months of nuclear power school, and then six months of hand-on hands-on prototype training before arriving in the fleet.
The Fantasy of the Self-Regulating Lab
The central debate on the show turned on Mark Zuckerberg’s recent memo pushing back against AI safety slowdowns. Zuckerberg argued that labs possess natural incentives to make models aligned because nobody wants to use misaligned agents. On the podcast, Coogan and his guests diagnosed this as a deliberate sidestepping of existential risk, contrasting Zuckerberg's soothing assurances with Anthropic’s willingness to grant independent watchdogs direct Slack and badge access. But looking at this through the ledger of actual industrial history reveals a deeper illusion. Silicon Valley wants us to believe that massive capital expenditures and proprietary model weights can police themselves through corporate goodwill. We have run this experiment before in financial services and rail transport, and the result is always captured regulators and deferred liabilities. A blog post about product safety is not an audit; it is a press release.
Capital Markets and the Compute Crackdown
The episode also took a sharp turn into macroeconomic reality with the Commerce Department reportedly forcing the shutdown of KHI’s AI compute price tracker over national security concerns. On the podcast, the hosts discussed how futures markets for Nvidia chips and data center capacity were starting to price the physical constraints of the AI boom. When the state steps in to freeze prediction markets for compute collateral, it signals that the underlying buildout has become too big, too leveraged, and too strategically vital to leave to open price discovery. Wall Street spent the zero-interest-rate era dreaming up software business models that vanished the moment money cost something; now that frontier hardware is backed by billions in debt, the state is clamping down on the scoreboard to prevent a panic.
The Reality Check on Payroll and Power
For all the talk of agent swarms and nuclear-level doom, the underlying architecture of this boom rests on physical steel, grid power, and capital costs that have just permanently shifted. The Federal Reserve raising rates by a quarter point to three and three-quarters and four percent confirms that cheap money is not returning to bail out speculative tech architectures. When energy shocks collide with data center power demands, the bottlenecks are no longer theoretical. If these frontier labs want to consume half a county's grid to train models that may or may not achieve recursive self-improvement, the public has a right to demand more than a pinky-swear from a billionaire executive. A technology that reorganizes the American economy needs an institutional framework tougher than a corporate comms strategy.
Policy is measured by what survives the first audit, not what sounds good on a morning talk show. Until these firms open their books and their server racks to independent verification with real teeth, every multi-billion-dollar cluster is just a high-stakes gamble waiting for the power bill to come due.