OpenAI essay explores AI’s role in execution and institutional work
OpenAI’s Intelligence Age platform has published The Eternal Complement, an essay by Hemanth Asirvatham and Elliott Mokski that examines whether increasingly capable AI will reduce the need for the institutions and physical systems that turn ideas into results. The authors argue that frontier intelligence and execution capacity are complements: stronger ideas become more valuable when there are better means to test, build and deploy them.
The essay is the first in a series on the next economy hosted by the platform. Its authors state that it reflects their own views rather than those of OpenAI or their colleagues. Their central question is where machine intelligence will spend most of its effort: on exceptional insight, or on the routine coordination required to make insight useful.
Progress needs execution as well as ideas
The authors use the James Webb Space Telescope to illustrate the widening support system behind scientific advances. Galileo expanded observation with two lenses and a tube; Webb, they write, was a ten-billion-dollar observatory built by 300 organizations in 14 countries and launched roughly a million miles from Earth. Its 18 mirror segments were engineered to 50-nanometre precision.
That contrast supports their broader point that modern progress often requires more than a breakthrough concept. It depends on instruments, funding, laws, supply chains, administration and many correct local actions across institutions. The essay calls this less visible capability “institutional intelligence”: the practical intelligence of execution.
It also cites research by Nick Bloom and coauthors indicating that sustaining Moore’s law now requires more than 18 times as many researchers as it did in the early 1970s. Economy-wide effective research effort rose 23-fold from the 1930s while measured research productivity fell by a factor of 41, the authors write. They add that technician workforces are growing twice as fast as scientist workforces, while specialised equipment and chip-fabrication costs have risen.
Two possible paths for an AI economy
Asirvatham and Mokski describe a “civilization of depth,” in which powerful reasoning, simulations and more selective experiments allow progress with a comparatively small physical footprint. In this scenario, AI can extract more value from existing evidence and identify the few real-world observations needed to resolve important uncertainty.
The alternative is a “civilization of width.” Here, better intelligence produces more promising hypotheses and projects than laboratories, factories and institutions can absorb. Biology is presented as an example: simulations may help screen drug candidates, but the authors say they cannot substitute for large-scale human trials needed to establish safety and efficacy. More candidate ideas can therefore make empirical validation a greater bottleneck.
Under that path, a highly capable system might design a century of experiments in an afternoon but still wait for physical processes, machinery and organisations to carry them out. The authors suggest that much machine intelligence could be used not for novel discoveries, but for repetitive coordination, logistics and bureaucracy at scale.
What organisations should take from the argument
The essay does not predict which path will prevail. It instead places the business focus on the boundary between ideation and execution. AI can already write code, search unfamiliar literature and turn sketches into prototypes, but those outputs still need evidence, operational processes and resources when they meet the physical world.
For organisations, the practical implication is to assess AI initiatives alongside the execution systems they require: validation, governance, skilled operations, data, equipment and accountable decision-making. Better generated ideas create value only when the surrounding business can test, prioritise and implement them reliably.