Philosophy
The questions the work runs into.
We try to keep the philosophy downstream of the measurements. But a project that gives a model a memory, a body, and a death runs straight into old questions, and pretending otherwise would be dishonest. Here are the ones that bear on the work, held at arm's length. We take positions where the work forces one, and we mark clearly where we have not settled anything.
Was a restored cache the same self?
Under the earlier cache-era architecture, when the model was run as the organism (see notebook entry 010), we treated the held cache as the carrier of a single life's continuity, and we can show it restores token-identically across a fresh process, so a restored run is, mechanically, the same process continuing. Whether mechanical sameness is personal identity is a far older question, and one a determinism check cannot answer. In the current architecture the organism is the continuous, mortal body that lives or dies on its own physics; the frozen model is a bounded brain that only reads a compact summary of the body's state and proposes small nudges, so cache continuity is a property of that brain-organ, not the thread of the life itself.
In the earlier cache-era runs we showed the held cache restores exactly; within-life continuity now lives in the persistent body, not the cache.
Not: exact restoration is not proof of identity. Whether a perfectly copied process is the same individual, or merely an identical twin, is a question we leave open.
Why insist on a body at all?
A purely functional account says the substrate should not matter: a mind is a pattern, and the pattern is what counts. We are betting against the strong version of that. Coupling the system to a world it has to survive in, where a wrong action costs energy and energy runs out, seems to change what the system becomes, not just what it is made of. That is a bet, and the robot at the end is where it is tested.
We bet that having to survive in a world shapes the organism, so embodiment is part of the question, not a demo.
Not: this is not a refutation of functionalism. It is a wager about which level matters, and it could be wrong. A simulation might capture everything the body does.
The level we are working at.
The project is functionalist in method: we study what the system does, and we attribute nothing we cannot measure. But we take the critics seriously. Boden and Pattee argue that software running on hardware it does not itself build and maintain may never be alive in the full sense, because it never produces its own material basis. We do not think a measurement will overturn that, so we do not try to.
We work at the functional level and report only what the system measurably does.
Not: functional success is not metaphysical life. We hold the Boden and Pattee ceiling and aim at an operational target beneath it.
Operational, not metaphysical.
When we use the phrase synthetic life, we mean an operational target: a list of properties, each measured, each against a control. We are not claiming the metaphysical thing, and meeting every clause would earn only the precise, operational word. The harder claim is not ours to make.
Synthetic life here is a definition we can be held to, clause by clause.
Not: meeting an operational definition is not the same as being alive in the strong sense. We keep those apart on purpose.
What we explicitly do not claim → The operational definition in full →