AI could give more people the means to pursue ideas, but a flood of machine-generated discoveries may leave the world short of the institutions, experiments and infrastructure needed to act on them. That is the central tension in an essay by Hemanth Asirvatham and Elliott Mokski, who argue that intelligence and the capacity to put it to work are complements.
Watch Desk analysis
What happened
The essay sets out two possible paths. In a “civilisation of depth”, better reasoning and simulation could help researchers get more from each real-world experiment. In a “civilisation of width”, discoveries still depend on increasingly elaborate experiments, machinery and institutions. The authors say history so far points towards rising demands for that supporting capacity, but argue that smarter complements could change the pattern.
They suggest AI is already making execution less scarce by helping with tasks such as writing code, searching literature and turning sketches into prototypes. That could let individuals pursue work once requiring a larger organisation. But if AI also produces many more promising ideas, the bottleneck could shift back to execution: deciding what is worth building, then finding the people, funding and physical resources to do it.
The authors describe this as their own view, not OpenAI’s or their colleagues’. The essay appeared on OpenAI’s platform, which hosts independent voices on an AGI future.
Why it matters
The familiar question is whether AI can produce clever ideas. This essay asks what happens after that: who can test them, build them and decide which deserve scarce resources? That puts laboratories, supply chains and institutions into the AI story, rather than treating progress as a contest of brainpower alone.
It also offers a useful distinction for readers weighing AI’s promises. More ideas could mean wider access to creative work, but it does not automatically mean more discoveries tested or more useful things made.
Our read
This is a thoughtful argument, not a forecast with the future neatly stapled to it. Its strongest point is that intelligence alone does not make progress; ideas still need contact with the world. If AI makes execution cheaper, that is exciting. If it makes worthwhile ideas arrive faster than society can test them, the old bottleneck may return wearing a new hat.
What to watch
- Whether AI-assisted research produces results that can be tested outside simulations.
- Whether smaller teams can turn prototypes into work that needs substantial funding or infrastructure.
- Whether institutions adapt to a larger volume of potential ideas, or become the constraint. Activity teaser: AI could help more people turn ideas into prototypes, but what happens if the ideas arrive faster than labs, institutions and supply chains can act on them? Hemanth Asirvatham and Elliott Mokski’s essay sketches two possibilities: a future where better reasoning makes each experiment go further, and one where progress still needs ever-larger efforts in the physical world. They present it as their own view, not OpenAI’s position. The useful question is not only whether AI can think up more. It is whether the rest of the world can keep up. Should AI’s biggest promise be making execution cheaper, or should we worry more about who gets to choose which ideas are worth executing?
Discussion spark: If AI makes it much easier to generate promising ideas, should society prioritise expanding the capacity to test and build them, or become more selective about which ideas get pursued?
Sources and evidence
- The eternal complement – openai.com (4 October 2026, 02:51 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.