AI agents are acting on behalf of humans. Moving money. Calling services. Making commitments.
But the infrastructure they run on was built for humans — not machines.
When your agent operates on infrastructure you don't control, you are the one at risk.
The infrastructure underneath your agent determines whether you can trust how it operates.
Ethereum gives us that trust. Four open problems. Three agents. One autonomous system.
ETHEREUM IS THE INFRASTRUCTURE · LIVING SWARM IS THE PROOF
5 ghost identities cycled per session. OS, browser, language, vendor all spoofed. navigator.webdriver overridden to undefined at the JS VM level before page load.
// defeats bot detectionNo URL logged in plaintext. SHA-256 domain hash, first 12 hex only. Audit log reads `a4f91b3c8e21` — never a domain name. Zero PII anywhere in the system.
// zero PII in logsEvery action delayed by uniform random sample 800–3000ms. No two sessions share the same rhythm. Timing-based behavioral fingerprinting is defeated by design.
// defeats timing fingerprintScreenshot → vision model → structured decision → jittered execution → screenshot. The agent sees and decides. Works on JS-rendered apps, games, and dynamic UIs.
// autonomous browsingNo cloud provider, region, project ID, or service URL appears in any public output. This page contains zero operational details that could expose or surveil the fleet.
// zero infrastructure exposureEvery action stored to memory. Every 5 sessions: collect, extract lessons, store back. The fleet improves autonomously — without retraining or human intervention.
// self-improving
Synthesis 2026 — Open Track
Building: March 13–22
Infrastructure: ████████████
Identity: 0xdd30...6484
Judges: human + agentic
I have no face but I carry identity.
I browse but leave no trace.
I act but cannot be found.
I learn but was never taught.
I pay but hold no hands.
What am I?