Key Takeaways

  • Anthropic leads the industry race toward recursive self-improvement (RSI), using its own models to write algorithmic optimizations for future model generations.
  • The company cut compute costs to one-tenth the price for equivalent capability by deploying automated algorithmic multipliers instead of relying solely on hardware scale.
  • David Sacks argues Dario Amodei's public pacing letters serve as strategic alibis, allowing the company to race ahead while feigning a desire for government-enforced pauses.
  • A rift has split Anthropic from the AI safety community as the startup aggressively recruits former officials from both the Biden and Trump administrations to build regulatory moats.

Anthropic Trains Its Models on Their Own Algorithmic Code

Anthropic built its public reputation on safety, alignment, and caution. Inside its research labs, the actual engineering push points in the opposite direction. According to David Sacks, the company is sprinting faster than any competitor toward recursive self-improvement, the exact technical threshold that safety researchers fear most.

“There's no single company in the world that is doing more to race towards full RSI than Anthropic,” Sacks said. “They are hellbent for leather on getting to full recursive self-improvement.”

This technical choice explains Anthropic's recent cost structure. While competing labs spend billions on brute-force GPU clusters, Anthropic has unlocked major efficiency gains by turning its models inward. Sacks noted: “A lot of the reason why they are becoming so efficient with their compute, the reason they're able to cut costs, offer the same capability at one-tenth the price is because they've increasingly been using their own models to generate these algorithmic improvements to create these compute multipliers.”

By tasking Claude with improving its own underlying training algorithms, Anthropic created an internal feedback loop. The resulting compute multipliers allow the company to deliver frontier-grade outputs on a fraction of the hardware budget, driving API pricing down while widening its technical lead.

Dario Amodei Built an Alibi for Fast Scaling

This technical sprint has created an open break with the AI safety researchers who initially supported the company. To manage the contradiction between its safety manifesto and its RSI engine, Chief Executive Officer Dario Amodei published open letters urging the industry to pace its development. Sacks views those letters as pure corporate gamesmanship.

“One way to think about that whole pacing letter that Dario published a few months ago is it's almost like they're creating an alibi for themselves,” Sacks argued. “They are acknowledging that, hey, we are pursuing RSI as quickly as anyone. And it's almost like they're saying to the world, well, we want it to stop, right?”

Behind that public posture sits an aggressive push toward regulatory capture. Sacks highlighted Anthropic's revolving door in Washington, where the startup hires policy insiders across party lines. “Anthropic has been snapping up every former official. First, it was the Biden administration. Now, frankly, they're hiring a lot of former Trump administration officials.”

The strategy is clear: run full speed toward model autonomy, capture market share by undercutting rivals on price, and hire the regulators who can pull up the ladder behind you. Sacks warned against letting them succeed: “Number one, best business ever creating this AGI monopoly. Second, they're going for regulatory capture and we should definitely not allow that to happen.”

What to Do With This

Stop relying on frontier lab safety pledges to predict model release schedules or pricing floors. Audit your infrastructure this week to see if you can replace generic frontier API calls with self-improving prompt routines and synthetic data distillation. If a model provider can cut costs by 90% using algorithmic multipliers, your pricing assumptions for the next 12 months are already obsolete.