AI could make it easier to turn ideas into working projects, but a flood of new ideas may leave laboratories, infrastructure and institutions struggling to keep up. That is the central tension in an essay by Hemanth Asirvatham and Elliott Mokski, published on OpenAI’s website as part of a series on the next economy.
Watch Desk analysis
What happened
The essay argues that progress depends on more than intelligence: experiments need equipment, funding, supply chains and people able to carry them out. It cites research suggesting that sustaining Moore’s law now takes more than 18 times as many researchers as in the early 1970s, and that measured research productivity across the US economy fell as research effort expanded.
The authors suggest AI may ease that execution bottleneck by writing code, searching literature and turning sketches into prototypes. But if AI also produces more hypotheses and designs, the bottleneck could move rather than vanish: there may be more promising work than the physical world can test.
Why it matters
The argument shifts the question from whether AI can produce clever ideas to what happens after the idea arrives. Scientific experiments take time; machines need to be built, and institutions have to organise the work. Faster thinking does not automatically mean faster discovery.
The authors frame two possible directions: a “civilization of depth”, where better reasoning and simulation make fewer real-world experiments necessary, and a “civilization of width”, where progress still needs ever more elaborate practical work.
Our read
This is a useful counterweight to the idea that smarter models alone will make progress run on autopilot. AI may lower the cost of getting from a hunch to a prototype, but the essay’s sharper question is whether it will also help choose which prototypes deserve scarce time, money and machinery.
The authors explicitly present the essay as their own views, not OpenAI’s. Treat its figures and forecasts as the case they make, rather than a settled account of what AI will do.
What to watch
- Whether AI-assisted research produces measurable gains in experiments completed, not just ideas proposed.
- Whether simulation can reliably narrow which real-world tests are worth running.
- Whether funding, equipment and research teams can keep pace with any increase in candidate projects.
Discussion spark: If AI makes it cheap to generate far more promising research ideas, should the priority be building more capacity to test them or making tougher choices about which ideas deserve a real-world experiment?
Sources and evidence
- The eternal complement – OpenAI (1 October 2026, 17:09 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.