Recently, OpenAI published an AI-generated proposed solution to the Navier–Stokes problem, one of the seven Millennium Prize Problems announced in 2000, tackling a question that had been open for roughly 90 years.11. OpenAI, “On the Navier–Stokes Millennium Prize Problem,” September 8, 2026. ↩ Its agents arrived at the solution in about 88 hours.22. Around the same time, Tristan Buckmaster (NYU) and Levent Alpöge (Anthropic) produced related AI-assisted breakthroughs for Euler and other fluid equations. ↩ This was fast enough to provoke a mini-existential crisis in the mathematics community, with several famous researchers commenting on the event.33. Kai Williams, “Math Can’t Go On Like This,” The Atlantic September 15, 2026. ↩ What will happen to genuinely human research if most research becomes automated?
One analogy I like is bread (cheese probably works too). Back in France, the baguette tradition was my favorite bakery item, and it really felt worth the extra price compared to supermarket bread. Despite the name, and I know some of my French friends did not know this, its formal designation is actually fairly recent: it was defined by a 1993 French decree (!), as a way to distinguish this kind of bread from more industrial alternatives.44. République française, “Décret n° 93-1074 du 13 septembre 1993,” Journal officiel de la République française September 14, 1993, article 2 (pain de tradition française). ↩
Maybe we will need something like the baguette tradition for research too. It is tempting to define “artisan research” as research done without AI, but artisan bread isn’t made without machines. Bakers use mixers and ovens; what matters is the ingredients, the process, and the “attention” given to the result. The equivalent in research might be that the researcher stays closely involved by choosing the question, developing taste about what matters, and ultimately taking responsibility for what they believe.
The analogy also works for training. Baguette-making is passed down through apprenticeships;55. UNESCO, “Artisanal Know-how and Culture of Baguette Bread,” Representative List of the Intangible Cultural Heritage of Humanity, 2022. ↩ keeping some research artisanal could similarly make sure humans still learn how research is actually done and how to judge it, instead of eventually delegating the whole craft to machines. For that to work, we would need credible community standards for what counts as artisan research, and people willing to value it. Like its bread counterpart, “supermarket” research can still exist, of course, and have its own customer base.
What is nice about this analogy is that it gives some optimism without requiring us to believe humans will always be better than machines at research. Machines can already make bread, and they may eventually make extremely good bread. Yet, we still care about bakeries; at least the one I buy from. Perhaps research could become similar: automated research becomes abundant and cheap, while there remains a place for work where someone has spent a long time caring about one question, and you can somehow feel that in the result.