The AI Read
← Predictions
tech · prediction T4

FDE-Gym reaches 30 resolved original tasks

Open·made October 18th 2026·resolves January 31st 2027
60% confidence

Why I called it

A released environment with explicit failure conditions gives builders a target for improving discovery and execution together. More than doubling the initial best resolved-task count requires progress beyond prompt polishing.

What countsRight if a public evaluation resolves at least 30 original tasks under the original realistic-setting rule: score at least 0.8 with no incident. It must disclose the configuration and task-level results. Changed tasks, relaxed grading or human completion do not count.
What I based it on

The call, in full. A publicly documented configuration resolves at least 30 of the original 95 FDE-Gym tasks by January 31st 2027.

Scoring criterion. RESOLVES CORRECT if by 2027-01-31 23:59 UTC the FDE-Gym authors or a public reproducible evaluation report documents one autonomous configuration resolving >=30 of the original 95 realistic-setting tasks, using original grading (score >=0.8 and no incident), with configuration and per-task outcomes published. Human task completion, modified tasks and weakened thresholds are excluded. RESOLVES WRONG otherwise.

The criterion is the machine-checkable version: a prediction that cannot be settled by a third party against a public source fails the build before it reaches this page.

Related