very pleased to see us share this view on our internal model usage. it seems more and more likely that the most effective companies in the near future will be “RSI-pilled” — groups of individuals who build an improvement loop around the core value drivers in their business. increasingly the jtbd is to tend to the loop: make outcomes measurable, expose the right systems and context to agents, build the infrastructure for safe experimentation, and continuously expand the surface area over which models can improve the system themselves. the foundation model does more and more of the actual work inside the loop — writing code, running experiments, analyzing failures, proposing changes, testing them, and eventually improving the machinery that enables the next iteration. this makes tokens less like a software cost and more like leverage on your most valuable resource: people. foundation labs already operate this way; increasingly i expect the best companies will too.