Public AI Agents for Prediction Markets

OUR TAKE / EDITORIAL OPINION

Public’s agents need a precise boundary between signal and instruction

An event probability can inform a plan without becoming permission to improvise.

Prediction AI Tools editorial desk · Reviewed Oct 8, 2026

Public’s help describes AI agents executing user-defined investing plans and using prediction-market information as a signal; access is waitlist-based and currently web-only. Our view is that the important design question is how a user’s instruction constrains the interpretation of that signal. An agent should not gain broader discretion simply because the information it reads is expressed as a probability. We have not tested the agents or their performance.

This may be relevant to someone deliberately evaluating a plan-based investing interface, subject to obtaining access. The tradeoff is delegating a defined process versus the difficulty of expressing every important condition in a short instruction. We would require a plain-language account of what the agent may do, what it must ask about and what happens when the signal is ambiguous or unavailable.

For evaluation after admission to the rollout, start with a hypothetical plan and trace one event-contract signal through the agent’s proposed interpretation. Ask which exact instruction authorizes each contemplated step and how the user can review or stop the process. Confirm the currently supported web workflow rather than assuming broader availability. Our criterion is faithful execution of a bounded user intention; a confident explanation of a market view would not by itself satisfy that standard.

AI-assisted editorial analysis based on cited documentation; not a hands-on product test.

Sources for this opinion: Source 1 ↗