February 2026, Truth Warrants paper at IC2S2
Our paired SSRN/IC2S2 papers on market-design interventions for safer agentic AI are now public. Here's the one-paragraph version.
If you build accountability into the marketplace where agents act, the rest of the system gets safer almost for free. That's the thesis of the SSRN preprint on market-design interventions for safer agentic AI, and the companion IC2S2 paper on interactive marketplace experiments for platform governance.
The companion paper shows that marketplace-style experiments are a better tool for testing moderation policies than offline counterfactuals. We ran the experiments inside Arbiter; the protocol is documented in the appendix.
The link rolls from this issue (Builds, bugs, and breakthroughs / Cool research to follow / Useful data drops) are indexed in the SimPPL Library and tagged with the source attribution we keep on each entry.