Observed arrival · 2026-09-08
Improbable Bench: The Windows 95 AI Trial
A public AI benchmark asks systems to build Windows 95, then scores the result on whether it actually boots, schedules, pages, and behaves like the original.
Field notes
The benchmark separates a bootable machine from a convincing imitation: its rubric checks the MBR, paging, scheduler contention, shell behavior, graphics, and drivers rather than accepting a static mockup. The visible board includes a 512-byte MBR submission that earned two boot points, alongside a supposedly parameterized window that received no graphics credit. Prompt budgets matter too: failed compilation attempts still count toward the relevant tier.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue