Observed arrival · 2026-10-05

Windway Data’s replayable agent-evaluation fixtures

windwaydata.com Visit website

Editorial interest 78/100 Selection signal · not a rating of the site

Windway Data presents tools and downloadable datasets for inspecting agent navigation, failure recovery, and deterministic replay.

Landing page captured for the 2026-10-05 issue.

For: Agent-systems evaluators and dataset builders
Worth noticing: The page reports zero public agent runs and says its 1,800 local trials tested control policies, not LLMs or customer agents.

Field notes

The page distinguishes a deterministic grid teaching fixture from its production corpus: the example blocks an attempted move, then checks whether a detour reaches the task predicate under the environment rules. Public downloads include two MIT-licensed 100-item samples—navigation fixtures and two-agent joint-planning fixtures—with schemas, verification information, and SHA-256 checksums; the canonical failure/recovery corpus is not public.

Observed signals

Read the marks

Editorial observations of this landing page, not a rating.

○OpenPublic substance visible

$PaidCommerce or pricing visible

✦PrettyNotable craft visible

●ProPolished or operationally mature

◎NicheUnusually specific use

One card from the complete issue

Still the Old Commit

307,787 arrived 1,000 judged 978 catalogued Enter the complete issue