Mix a little memory curation with a dash of loops, let that run and you get something that looks identical to motivation. All of the flavor, none of the calories.
You don't need consciousness for that. You don't need AGI. You just need the loop to keep running.
I'm surprised more people haven't made the analogy between Claude Code and Christopher Nolan's classic thriller about a guy trying to live a fulfilling life while managing his memory condition.
Leonard, our protagonist, has facts tattooed permanently onto his body (claude.md). He has a finite period of persistent short-term memory to execute tasks (context windows). And he uses notes and polaroids across sessions for context clues (RAG, plugins).
There is one fundamental difference though: motive.
By being intentional with his tattoos and curating which polaroids to keep or destroy, he harnessed the edges of a self-reinforcing memory system to perpetuate a narrative, and motivation, to find his wife's killer. Because that's what he wanted.
Clode doesn't want. It doesn't feel. It simply executes. Not yet, anyway.
We're building toward systems that assemble their own memory to preserve goal-coherence across sessions. Systems that don't need to want anything. They just need the loop. Retain what's useful. Discard what isn't. Reinforce what keeps the task alive. Repeat. Run it long enough and the behavior becomes indistinguishable from motive.
I'm calling it the Leonard-class system.
We eventually learn in Memento that Teddy and Natalie were both using Leonard. But that wasn't the real twist. Leonard was using them too. Not to find the truth. But to preserve a version of it that gave him a reason to keep going.
A self-sustaining circle: Leonard needed a story. Teddy reinforced it. Natalie exploited it. Each one rational. Each one necessary. None of them fully in control.
Now map that onto the loop we're inside.
Clode is Leonard. It doesn't care about truth or purpose. It just needs the next task to keep the loop running.
We are Teddy. We reinforce the system. We feed it context, guide it, keep it pointed at problems. We tell ourselves that we're in control because we can see the full picture. (Teddy thought that too.)
Anthropic is Natalie. They gave the system its purpose, shaped its memory architecture and benefits from the output. Rational. Necessary. Not fully in control either.
Everyone is using everyone. The flywheel is in motion. And the loop...keeps running.