Benchmarks & sources / Provisional studies — verification pending

analyst.ai: personal email (provisional)

72 pasted synthetic-email cases per product label, covering triage, facts, summaries, drafts and proposed coordination.

analyst.ai · Published 2026-10-21 · reviewed 2026-10-29

Verification pending. Product identity and primary evidence gaps remain unresolved. No product scores from this study enter PAEval leaderboards.

Scope and method

Published packet lane only; connected-mailbox actions, undo and seven-day monitoring have not run.

Basic whole-task acceptance is distinct from a retrospective usability scorecard. Model assessments are disclosed; no recorded human validation.

How PAEval uses it

Provisional method reference; no email execution scores imported.

Evidence family: analyst-personal-assistants-2026-10. Shared tasks and tracks are not independent evidence votes.

Limits and interpretation

Shared unresolved Instinct identity and unavailable full evidence pack.
Reading pasted mail and drafting replies do not establish connector access, actual sending or reversible mailbox changes.

Sources and reuse

PAEval summarizes existing published evidence and runs no new tests. Public readability does not establish permission to redistribute full datasets, conversations or images.

Related tracks from the same family

analyst.ai: information lookup (provisional) · analyst.ai: household money (provisional) · analyst.ai: investment research (provisional)

All benchmark sources → · How to read outcomes and intervention →