When I first read how Kite structures identity, I caught myself asking: “Why so much ceremony just to pay for something?”
Authority is split into three layers. The user holds the root, the agent gets delegated authority, and each session runs on a temporary key. The simplest agent payment setup skips all of that: hand the agent a wallet and let it spend. If the goal is autonomy, fewer layers looks like the obvious design. Measured by capability, it is.
But I don’t think Kite is competing on capability. What it’s pricing is Blast Radius.
An agent’s judgment is the one thing nobody can guarantee. Models drift, prompts get hijacked, a tool returns something odd and the agent acts on it. Capability keeps improving and the failures keep arriving anyway. So the question that matters for an agent holding money isn’t how much it can do. It’s how much damage its worst day can do.
Seen that way, the Authority Chain does a different job. The agent never touches the root key. A session never carries the agent’s full authority. The design keeps a breach inside the layer where it happened, and Passport enforces spending limits on top. The worst case stops being a surprise and becomes a number someone chose.
That’s why I read Kite as shrinking mistakes rather than preventing them.
There’s a catch, and it’s the part I keep chewing on. A small Blast Radius only guards against what the limits forbid. An agent talked into overspending inside its budget, at a merchant it’s allowed to pay, passes every check. The key stayed safe. The money still left.
So the radius is a setting, not a guarantee, and settings drift wide when people want convenience. The evidence I’d weigh is how tight Kite’s defaults are and how many users loosen them.
If most do, the three layers protect keys. If most don’t, they protect money. Which one will Kite’s users end up with?
#KiteBBW $KITE 😇
Authority is split into three layers. The user holds the root, the agent gets delegated authority, and each session runs on a temporary key. The simplest agent payment setup skips all of that: hand the agent a wallet and let it spend. If the goal is autonomy, fewer layers looks like the obvious design. Measured by capability, it is.
But I don’t think Kite is competing on capability. What it’s pricing is Blast Radius.
An agent’s judgment is the one thing nobody can guarantee. Models drift, prompts get hijacked, a tool returns something odd and the agent acts on it. Capability keeps improving and the failures keep arriving anyway. So the question that matters for an agent holding money isn’t how much it can do. It’s how much damage its worst day can do.
Seen that way, the Authority Chain does a different job. The agent never touches the root key. A session never carries the agent’s full authority. The design keeps a breach inside the layer where it happened, and Passport enforces spending limits on top. The worst case stops being a surprise and becomes a number someone chose.
That’s why I read Kite as shrinking mistakes rather than preventing them.
There’s a catch, and it’s the part I keep chewing on. A small Blast Radius only guards against what the limits forbid. An agent talked into overspending inside its budget, at a merchant it’s allowed to pay, passes every check. The key stayed safe. The money still left.
So the radius is a setting, not a guarantee, and settings drift wide when people want convenience. The evidence I’d weigh is how tight Kite’s defaults are and how many users loosen them.
If most do, the three layers protect keys. If most don’t, they protect money. Which one will Kite’s users end up with?
#KiteBBW $KITE 😇
