Science fiction for you for the night. Agency economies for autonomous scientific discovery.

Science fiction for you for the night. Agency economies for autonomous scientific discovery.

Nenad Tomashev, Mathia Franklin, Atusa Kasirzada, Vivek Natarajan, Alan Karthikslingam, Sebastian Carrier, Simon Osindro.

paper:. review:. Code: Not available.

Model: Not available.

What they did: The authors introduced agency economics for science, a market architecture for autonomous scientific systems (multi-agent models that generate hypotheses and conduct experiments). It coordinates scientists, artificial intelligence agents, and robotic labs: sets prices for data access, enables trading in token research applications, and validates physical experiments through smart contracts. What we got: In this conceptual paper, the authors note that the cost of generating computational hypotheses tends to zero, so scientific inquiry is limited to real-world inelastic tests. For example, according to Iroma's law, the cost of drug approval by the FDA doubles every 9 years and has already reached $2.6 billion per molecule. To weed out weak ideas before the lab, the framework introduces a 4-layer data mart and a 4-step hypothesis exchange pipeline that separates idea generation from wet lab experiments. Why it matters: The work shifts the focus of AI for science from isolated thought enhancements to explicit cost accounting, combinatorial resource auctions, and automatic depositing for replication of results. However, the entire proposed infrastructure remains conceptual and does not yet have an empirical prototype. Get flowing with ideas here: