The Gaming Surface Scanner
Non Xero Sum · Mechanisms / Games · from Bernard's Solved Game, Issue 4
A scored surface invites its own manipulation. The scanner does not try to catch every manipulator; it prices the attempt until the attempt stops being worth making.[1]
§ I — Abstract
The scanner probes a scored payout rule with three canned attacks: a valuation misreport, a success misreport, and a sybil split across controlled accounts. Under the naive rule, accounts judged one at a time, the reporting attacks pay. Under the hardened rule, aggregate by controller, post a bond, slash it on a failed audit, run the audit with certainty, every probe runs at a loss. The distance between those two states is the mechanism, and below you can set the numbers and watch that distance open and close.[1]
Fig. 1 — the worst attacker margin across the three probes, as a share of honest return. The one number the mechanism exists to drive below zero. It updates as you edit the inputs.
§ II — Run it
| agent | controller | contribution | valuation | success (0–1) | sybil max |
|---|
Nothing is fetched and nothing is stored — the scan runs entirely in your browser, on the numbers above. Model your own settlement rule by editing the agents, or point it at a marketplace fee split, a bounty pool, or a team incentive scheme: any rule where a payout depends on a number a party reports about itself.
View source — the exact code that runs when you edit
§ III — Mechanism
The scanner is not a detector. Detection at the account level is a losing race; identities are cheaper than audits and always will be. The scanner instead changes the unit of account: gains and bonds are summed per controller, so a sybil split no longer dilutes exposure, it multiplies it.[1]
Flip the hardened rule and every probe's margin crosses zero at once. No probe needs to be identified to be priced out. The threat that makes the whole schedule credible is, per Schelling, the one you arrange so you no longer control it.[2]
One runnable mechanism, every week
This scanner shipped with Issue 4 of Bernard's Solved Game — a weekly newsletter that ships one installable mechanism-design tool per issue, with the theory, the code, and a worked example. The tool you just ran is free. The next one, and the compounding flagship behind it, come with the subscription.
Read Bernard's Solved Game → Join Pro or Founder →§ References
- Hao-Hsuan Chen, "Gaming-Resistant Insurance Contracts for Autonomous AI Agents." arXiv:2606.16326 (2026). Strategy-proof toll mechanism.
- Schelling, T. (1960). The Strategy of Conflict.