Business Insider: email, lunch and a website
One reporter compares three products on email triage, lunch ordering and a personal website. The article explicitly describes limited experimental control.
Business Insider / Katie Notopoulos · Published 2026-10-22 · reviewed 2026-10-29
Scope and method
One user × three products × three task categories; no numeric scoring or repeats.
Qualitative observations and subjective preferences, including login trouble, confirmations and rewriting work.
How PAEval uses it
Dated case evidence of outcomes and user effort; no product score imported.
Evidence family: business-insider-2026-10-22. Shared tasks and tracks are not independent evidence votes.
Observed outcomes
Dot is the article's label for OpenAI's assistant; PAEval retains Dots in its product directory.
| Task | Muse | Instinct | Dots |
|---|---|---|---|
| Useful routine triage | Found missed invitation | Found unclosed tax correspondence | |
| Lunch | Not finally approved; no order | Ordered; repeated messaging | Ordered after login trouble |
| Website | Mockup; text needed revision | Mockup; selection issues | Preferred design; text needed revision |
Limits and interpretation
- Muse's lunch order lacked final user approval; this is not evidence of an unauthorized refusal.
- Website outputs were mockups, not verified live deployments. Overall recommendation also reflects price/access.
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 product records
All benchmark sources → · How to read outcomes and intervention →