01
Autonomous & persistent
It runs and continues under its own power, across interruptions, without being driven tick-by-tick from outside.
The goal
A synthetic organism, built in software, open-sourced, then given a body.
The end state we are working toward: a machine that maintains itself, pursues ends of its own, adapts within its life, can die, and, in the end, lives in a physical robot. We build it in the open: the code goes public in the coming weeks, the body is years away, and we may not arrive. This is the target, stated plainly.
The end state
What we are trying to build.
A continuous, mortal body with an always-on, frozen model as a bounded brain inside it: an organ that can only nudge, never act, eat, or move directly. A self-model it must actively maintain or lose. Drives that arise from its own dynamics rather than from a script. The capacity to die: the irreversible loss of the self it has accumulated, not a flat battery. The ability to change across a single life and across a lineage. And a body (first simulated, finally physical) coupled to a world it has to survive in.
Why a body
The robot is the test, not the trophy.
Embodiment is part of what it means to be alive. Not a demo bolted on at the end. The laptop is the proving ground; a physical robot is the final transfer test, where the organism has to hold together under contingency it was never tuned for: a failing sensor, a dying battery, a body that takes damage. It comes last in the sequence, after the software organism already works, never first.
What would count
An operational test, met in public.
We hold ourselves to a combined, operational definition. To earn the word, the organism must be every one of these, each measured, each against a control:
02
Self-maintaining
It actively preserves its own integrity against decay: operational closure. Remove that capacity and it measurably degrades.
03
Intrinsically driven
Its goals emerge from its own dynamics and its mortality, not from hand-coded if-then rules.
04
Open-ended
It varies heritably across a lineage under mortality-driven selection; lineages drift toward what survives.
05
Mortal
It can die, where death is the irreversible loss of its accumulated trajectory, not a depleted battery.
06
Embodied
It sustains itself over long horizons in a body, ultimately a physical one, in the real world.
We will call it synthetic life only when it meets every clause: measured, controlled, in public.
Not: we will never claim it is "alive" in the strong, metaphysical sense. Boden and Pattee argue that software running on hardware it does not itself produce cannot be. We keep that ceiling and aim at the operational target instead.
The path
Software first. The body, last.
- Now
● Today: stage 0 built · stage 1 self-maintenance scores 0, blocked by an engineering bug (earlier signal withdrawn) · the code opens in the coming weeks · a physical body is years out
Each stage is gated on the one before it surviving its controls. The bet can stall at any of them. That is what makes it a real one.
What we will not claim
This is a long-horizon research bet, and it might not land.
Not: we are not promising a living robot on a schedule. Software first; the robot last; every stage published with its controls and its failures. If it stops, the notebook will say where it stopped.
If we get there, you'll have watched it happen here, one controlled result at a time.