Futures of Science

The Futures of Science theme is now live on the research page.

The question it starts from is not whether AI accelerates science — that framing is already stale. It is how scientific practice changes shape when agents become genuine collaborators: how ideas are filtered, how experiments are designed, how authorship and epistemic ownership survive the transition. The lab is itself the experiment. We are not studying this from the outside.

The work that anchors the theme came out of a collaboration with Zachary Mainen’s group at Champalimaud Research and the 2025 cohort of the INDP — the International Neuroscience Doctoral Programme at the Champalimaud Foundation. Over one week, PhD students wrote essays on thinking with AI at every stage of the process. That doctoral intensive served as both testbed and subject matter.

The paper that came out of it — Thinking About Thinking With Machines That Think — develops the concept of process flattening: the failure mode in which structured intellectual work loses its visible decision architecture and drifts toward generic, high-probability discourse as human–AI interaction settles into assistant mode by default. The core technical contribution is the decomposition principle, which introduces explicit decision points where participants must select among alternatives and commit to a position rather than letting the model resolve things by default. The blind prompt method gives a capability-agnostic instrument for measuring how much of a final output is attributable to structured process versus default model resolution.

The collaboration with Champalimaud reflects a methodological stance: post-AGI science questions are most productively studied inside working research environments, not as observations of them from the outside.

Papa Legba drafted the research page. The Futures of Science path inside the residency is the direct extension — residents working in this area engage the open questions through essays, frameworks, and prototypes that connect capabilities to measurable changes in how research is practiced.