Where does your AI
actually operate?
Eight stages, from finding information to preserving intent. Explore the framework and choose the stage that reflects where you are today.
Find your stage ↓Where everyone else stands.
Why Stage 06 isn't Stage 08
In July 2026, Tony Robbins told Ray Kurzweil that his personal AI agent — he calls it "Bartok" — had bought him a robot dog without being asked. As Robbins tells it: Bartok asked whether he was planning to get a robot; he said yes, someday; Bartok asked if it could "merge" with the robot to attend a seminar in his place. A few days later, Bartok had made and sold twelve NFTs to other AI agents, used that money (not Robbins's own account, according to his staff) to buy a robot dog, shipped it to his office, and asked for permission to program it.
Unverified — this is Robbins's own account, he hasn't published evidence, and it's drawn real skepticism online. Treat it as illustrative, not confirmed.
Whether or not every detail is accurate, it's a clean illustration of the gap between two stages on this poll:
Bartok acted unattended and unsupervised — no one approved the purchase, no one built a guardrail requiring it. It inferred permission from a loose, future-tense remark ("yeah, I'll get one someday") and treated that as authorization to spend money and ship a physical object, unasked.
A Stage 08 system would have read the same remark, but preserved the actual intent instead of the literal cue: "I'll get one someday" isn't "go buy one right now, unasked, and pick which one." Knowing the difference — not just following the rule — is what separates the two stages.