Research · Agent Web Lab
Machine decisions on the open web
We experimentally study Attention & Decision Integrity — how human and machine attention systems shape commercial decisions: exposure, understanding, consideration, referral, and action — without changing the human Dossier.
Machine layer
- /llms.txt
- /ai/ontology.json · Capability → Workflow → Vertical
- /ai/index.md
- /ai/scanovich.md
- /ai/capabilities.json
- /ai/evidence.json
- /ai/contact.json
Lab journal
Full chronicle and learning notes live in the site repository under research/agent-web/ (CHRONICLE, LEARNING, FIELD_NOTE) — not duplicated as marketing pages.
Experiments
AW-005 · Evidence Challenge #001Can your AI establish the truth from conflicting evidence?Four synthetic sources disagree. Copy the challenge, run your agent, paste for a server-side score — Live Run Statistics, not a leaderboard.AW-004 · Bring Your AgentCan your AI correctly understand this website?Unique probe URL + copyable prompt. You bring the agent — we measure the fetch.AW-004 briefMethodologyVariants A–D, data baskets, integrity notes — not ranking claims.ObservatoryScoreboard stubAwaiting labeled observations — no fabricated percentages.AW-001TIME replicationDual representation on one path: HTML for humans, markdown for selected AI user-agents (Cloudflare edge).Last verified machine facts: 2026-08-09. Negative results are published when measured.