- Who it helps
- Teams testing automation in stateful desktop applications.
- Availability
- August 14 preprint revisited in the 90-day catch-up. It reports 28 Windows workflows and six hosted agents; public executable artifact access was not verified.
Inspect the final state
Atomicity means a task either completes correctly or leaves no unintended persistent effects. Initial states, goal states and task-specific validators make this distinction measurable.
LegacyWorld checks whether failed GUI tasks leave unwanted changes ↗
A failed run can still change records
The study treats useful completion, safe failure and partial harmful changes as different outcomes. A demonstration that reaches the right screen cannot establish all three.
LegacyWorld checks whether failed GUI tasks leave unwanted changes ↗
What you can do
- Read the state-validator definitions before designing acceptance tests for a comparable workflow. Open reference ↗
Where the evidence stops
Abstract checked October 31; no Windows application or benchmark was executed. Results remain bounded to the reported workflows.
Sources read. Sources checked October 31, 2026 UTC. No third-party software, model inference or complete workflow was executed.
Sources & further reading
Paper · paper
LegacyWorld checks whether failed GUI tasks leave unwanted changes ↗
Supports the cited capability or limit in LegacyWorld checks whether failed GUI tasks leave unwanted changes.
Thilo Reintjes and collaborators · page checked 2026-10-31
Coverage note: Independent attention and end-to-end execution remain unverified.
The record over time (1)
Dated source checks and published explanations. An observation date can differ from the original release date.
- · Added to Current
LegacyWorld evaluates whether a computer-use agent completes a workflow or fails without leaving unintended persistent changes.