Skip to content
Gaurav MishraHome

Ideas

A curated collection on operational AI, leadership, organizations, and the human systems that make autonomy useful.

A three-step safety test showing an agent recommendation, human review, and the question of whether the reviewer can catch a bad recommendation.
ObservationSystems & Autonomy5 min read

Human-In-The-Loop Isn't The Safety

A human approval step only contains risk when the reviewer can reliably detect a bad recommendation.

Read Observation
Three cards representing the enterprise AI bets that models will improve, people will adapt, and reliability must be engineered.
ObservationEnterprise Trajectories5 min read

The Three Camps of Enterprise AI

Three different bets—model progress, human adaptation, and reliability engineering—and what enterprise teams should borrow from each.

Read Observation