mudpie

Agent readiness methods · dated pilot protocol

Score the evidence. Keep the gaps.

A score is only useful when its denominator, exclusions, observation conditions, and missing work remain visible.

Scope

This first public scorecard is one scored pilot for Egoist Machines (yc:32129), observed 2026-09-20–2026-09-21 UTC in a logged-out, public, read-only scope. It is a current snapshot, not a cohort prevalence estimate, and the broader report set remains ongoing.

How the score is calculated

The four dimensions carry fixed weights: SEO 40%, AEO 25%, agent 25%, and reliability 10%. A pass earns all of a check’s points, partial earns half, and fail earns none. Within each dimension, the applicable denominator includes checks with pass, partial, or fail status. Earned points are divided by those applicable points, then multiplied by the dimension weight. The displayed score is the weighted sum: 85.1/100.

Score weights for this pilot
DimensionWeightEarned / applicableRate
SEO40%35.5 / 4088.8%
AEO25%19.875 / 2482.8%
AGENT25%20 / 22.2589.9%
RELIABILITY10%5.625 / 8.7564.3%

Applicability gates

  • N/A means the criterion is explicitly not applicable to this public surface. It remains visible and is excluded from the applicable denominator.
  • Owner-only is the public display label for the source status not_publicly_verifiable: the behavior may matter, but the logged-out public audit could not verify it. These points are excluded from the applicable denominator.
  • Unknown stays unknown. It is never turned into a fail, zero, or absence claim. This pilot has zero unknown rows after the reviewed sidecar, but later pilots may not.
  • A pilot can be scored only when its applicable check ledger and dimension denominator are complete. That score gate does not make the result a ranking or an adoption claim.

Observation boundaries

  • Lab Lighthouse values are lab observations. The mobile LCP deficiency is approximately 7 seconds in the saved run; it is not field telemetry.
  • Logged-out page reads, rendered accessibility checks, discovery documents, and bounded keyboard/mobile checks remain separate from registration, invocation, task completion, repeat use, and adoption.
  • No login, form submission, purchase, tool execution, rate-limit bypass, or security test was performed for this report.
  • The public export preserves all 87 original check statuses and the source score. The page maps N/A and owner-only to reader-friendly labels without changing the underlying export.

What this pilot does not claim

It does not claim that all 1,752 companies were graded, that one company is a category leader, or that a readiness score predicts traffic, citations, execution, conversion, or adoption. The native WebMCP census is a separate dated dataset and should be read on its own denominator.

Public artifacts

Source export SHA-256: 46db927b6ee23bd441d722a07e1870126de0267d36b917be64eed711ab939eda