Skip to content
Asante OS

Blog

The autonomy ladder: HOOTL, HOTL, HITL

A practical ladder for verified intelligence: inner loops that stay hands-off, outer loops with human oversight, and sensitive gates that require a human in the loop.

· 7 min read

Autonomy marketed as a single switch trains teams to confuse convenience with control. Asante OS uses a ladder: different rungs for different blast radii.

HOOTL — hands off the loop

Inner loops can run without a human in every step when verification is continuous and failure is cheap to reverse. HOOTL is earned, not assumed.

HOTL — human on the loop

Outer loops keep a human interruptible: oversight, sampling, and the ability to halt without being the bottleneck on every micro-decision.

HITL — human in the loop

Sensitive gates — irreversible actions, high-stakes commitments, ambiguous policy — require explicit human approval. That is a feature, not a lag.

Kill switch

If you cannot stop it, you do not control it. Haltability sits beside the ladder, not below the marketing fold.

← All posts