The autonomy levels

Teams use a ladder of levels to say how much of the work an agent does without a human. This page holds the two ladders the history uses, and the lesson both share.

Shapiro’s five levels (23 January 2026)

Dan Shapiro, CEO of Glowforge, published “The Five Levels: from Spicy Autocomplete to the Dark Factory.” He modeled it on the self-driving taxonomy. Source: Order 002.

Level	Name	What it means
0	Spicy Autocomplete	Not one character reaches the disk without human approval
1	Coding Intern	AI writes boilerplate and low-stakes snippets under full human review
2	Junior Developer	A human pair-programs with the model and reviews every line
3	Developer	Most code is AI-generated. The human is a full-time reviewer
4	Engineering Team	The human works on specs and plans. Agents do the work
5	Dark Factory	No human writes or reviews code. Humans define intent and review outcomes

Shapiro claims that about 90 percent of developers who call themselves “AI-native” stay at Level 2.

River’s driving-style levels (27 July 2026)

Order 001 uses the SAE driving levels. They ask who writes, who tests, who reviews and who decides scope.

Level	Name	Writes code	Runs tests	Reviews	Decides scope
0	Manual	Human	Human	Human	Human
1	Assisted	Human	Human	Human	Human
2	Partial	Human and agent	Human	Human	Human
3	Conditional	Agent, supervised	Agent and human	Human	Human
4	High	Agent	Agent	Agent, spot-checked	Human
5	Full	Agent	Agent	Agent	Agent, within mission

Other ladders in the record

Grab published a five-level model with four months of data (Order 023). China requires three tiers of decision authority before deployment (Order 016).

What the ladders agree on

The move from Level 3 to Level 4 is repository work. It is not model work (Order 001). Verification and governance gate the jump to Levels 4 and 5, not model capability (Order 002). Match the level to the stakes of the work. A repository can hold different levels for different directories (see Order 034 for autonomy by path).