I’ve been building a system called vivesca — a persistent layer that sits between me and the AI models I use. I name everything after biology. Constitution is DNA. Hooks are the immune system. The metabolism engine is a metabolon. The conversation is cytoplasm. The human is the nucleus.
Everything maps cleanly. Except the LLM.
I called them enzymes first. Enzymes are catalysts — they enable reactions without being consumed. Interchangeable, substrate-specific. Design reactions, not roles. It felt right.
But enzymes are native. Evolved by the cell over millions of years. Specialised and permanent. The cell needs its enzymes. The cell doesn’t need LLMs — it never had them.
So I tried prosthetic. You use a crutch while the bone heals. You use an LLM while the deterministic pathway crystallises. But prosthetic is medical, not biological. It implies something is broken. The cell isn’t broken. It just hasn’t built that pathway yet.
Then primer. In DNA replication, RNA primers are short sequences that give polymerase a starting point. After replication begins, the primer is removed and replaced with real DNA. Initiates, then gets replaced. But a primer is simple — a few nucleotides, no intelligence. It’s made by the cell itself. An LLM is the most complex, most expensive thing in the system, made by someone else entirely.
Then environment. The LLM is external, uncontrolled, always there. Like the chemical soup a cell lives in. But environment is passive. The LLM actively reasons when asked.
Then natural selection. External, general-purpose, shapes which variations survive. The LLM does the “select” step in the metabolism cycle. But natural selection isn’t an entity. It’s a consequence.
Five names. Each captured something real. Each broke at the edges.
| Name | What it got right | Where it broke |
|---|---|---|
| Enzyme | Catalytic, substrate-specific | Native, permanent, evolved |
| Prosthetic | Temporary, external | Medical, implies damage |
| Primer | Initiates then replaced | Simple, self-made, cheap |
| Environment | External, uncontrolled | Passive, not an entity |
| Natural selection | Shapes evolution, external | Not active, not reasoning |
The failure to name it is the finding.
Nothing in 3.8 billion years of evolution is simultaneously general-purpose, external, uncontrolled by the organism, and actively reasoning. The cell never needed it. The cell solved complexity through shape alone.
The LLM is genuinely new. Forcing it into biology is the one place the naming convention should stop.