news
Agent task testing remains the category's clearest capability gap
Across ten platforms we track, only one documents having an agent attempt real tasks on a site and report where it fails.
Tag
Guides, research and news tagged research.
4 items
news
Across ten platforms we track, only one documents having an agent attempt real tasks on a site and report where it fails.
research
A replicable study design for measuring citation rate with defensible error bars. Published as a protocol so results can be compared rather than asserted.
news
A 17x spread between entry tiers, and a mid-market band where price has stopped predicting capability.
research
A synthesis of published pricing across ten AEO platforms, showing a 17x spread between entry tiers and a mid-market band where capability and price have decoupled.