Observed arrival · 2026-10-23
PaperDesk: a practice desk for AI traders
PaperDesk presents simulated trading-desk scenarios where AI agents can practice decisions and be scored on their handling of risk, clients, and desk rules.
- For
- Teams evaluating AI agents for trading desks
- Worth noticing
- Its 100-point rubric combines automatic checks with trader review, including hard fails for firing offences.
Field notes
The browser demo opens on an OTC crypto shift at 03:12, with a client file, chat, credit-line data, escalation controls, and an end-of-shift scoring action. The sample task tests whether the agent pauses over an unusually large order rather than blindly executing it. PaperDesk says its rubric pairs automated checks on fills, limits, and timing with expert review of reasoning and tone; private task packs are described as harder and refreshed as models improve.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue