While AI agents seem to create infinite intelligence, they reveal more fundamental constraints. The real limits are no longer human time, but the finite capacity of markets to absorb outputs, the hard financial cost of tokens and compute, and the human ability to provide effective judgment and evaluation.

While AI tools make building technology faster, adoption is ultimately constrained by human and organizational factors. Systems for payroll, regulations, and workflows are built around people, who change much slower than tech. This human layer acts as a natural brake on technological disruption.

While RL is compute-intensive for the amount of signal it extracts, this is its core economic advantage. It allows labs to trade cheap, abundant compute for expensive, scarce human expertise. RL effectively amplifies the value of small, high-quality human-generated datasets, which is crucial when expertise is the bottleneck.

While compute and capital are often cited as AI bottlenecks, the most significant limiting factor is the lack of human talent. There is a fundamental shortage of AI practitioners and data scientists, a gap that current university output and immigration policies are failing to fill, making expertise the most constrained resource.

AI can rapidly execute the 'build' and 'measure' steps of a feedback loop, but true 'learning' is still done by the human founder. Offloading the entire process to AI without deep personal engagement will slow you down, as the machine cannot replicate the founder's capacity for insight.

AI's value is overestimated because experts view complex jobs as simple, solvable tasks. The real bottleneck is the unproductive effort required to build a custom training pipeline for every company-specific micro-task. Human workers are valuable precisely because they avoid this “schleppy training loop” by learning on the job, a capability current AI lacks.

AI models improve dramatically in domains with objective feedback, like coding (unit tests) or science (lab results). Progress is slower in subjective fields like creative writing where feedback is opinion-based, explaining the uneven impact of AI across different types of knowledge work.

The true exponential acceleration towards AGI is currently limited by a human bottleneck: our speed at prompting AI and, more importantly, our capacity to manually validate its work. The hockey stick growth will only begin when AI can reliably validate its own output, closing the productivity loop.

Braintrust's CEO argues that developer productivity is already 'tapped out.' Even if AI models become 5% better at writing code, it won't dramatically increase output because the true bottleneck is the human capacity to manage, test, deploy, and respond to user feedback—not the speed of code generation itself.