The Reaction Grammar and Reactor
What can happen when two atoms meet, what it costs, and the order in which it happens.
An artificial chemistry has three parts: the molecules that can exist, the reactions that can happen, and the algorithm that decides which reactions fire. The Chemistry supplied the first: twenty atoms, each with a valence, a polarity, a reactivity and a cost. This page supplies the other two. Everything here is authored by us and labeled as authored. Nothing here has been simulated yet.
One rule above the others: energy pays, only structure decides
The lab feeds the lattice two things, both entering at its edge, in a steady flow:
- Raw atoms. Single atoms, never already bonded to anything. Nothing arrives pre-assembled.
- Energy tokens. All identical, carrying no information, earmarked for nothing.
A token can pay for any reaction and can choose none. It never changes an atom’s type, valence or polarity. Cells work the same way: one energy currency is spent on thousands of different jobs, and the cell’s own machinery decides which. The rule matters because a supply that could become whatever a structure needed would be doing the structure’s work for it, and on the Ignition Watch that is an operator supplied from outside. Every token is counted in a balanced ledger of the kind the Modeled Economy already keeps.
The reaction grammar: seven events
This list is the whole of what can happen. An event not on it cannot occur in the lab.
1. Bonding
Two neighbouring atoms join if both have valence left and their polarities attract (opposite signs attract, like signs resist). It costs tokens, drawn from the cells involved. Holding a bond also costs tokens every tik; a bond whose holding cost cannot be paid breaks.
2. Breaking
A bond comes apart and both atoms get their valence back. Breaking neither costs nor returns tokens, so nothing in the lab can run as a free engine by making and breaking the same bond.
3. Catalysed bonding
A catalytic atom (k1 or k2) arrives as raw supply, like any other atom, and on its own it does nothing. It becomes a catalyst only when both of its bonds are to binding atoms (b1, b2 or b3): a binding site that a structure has built around it. A working catalytic site halves the cost of a bond forming next to it, reaches its immediate neighbours only, and is not used up. The working catalyst is therefore always something a structure made, which is what the Ignition Watch’s first test asks for.
4. Donor-to-acceptor transfer
A donor atom (d) passes one energy token to a neighbouring acceptor atom (a). This lets a structure move energy from where it arrives to where it is spent: a small metabolism, run by the structure’s own parts.
5. Decay
Every bond has a chance each tik of breaking on its own, larger for more reactive atoms. Decay is what makes a boundary wear out and have to be rebuilt. Without it nothing would ever need upkeep, and the Ignition Watch’s third test, rebuilding after a full turnover, could never happen.
6. Template copying
A bonded chain can act as a template. A free atom lying beside a chain bonds more readily to its neighbour on the far side when its polarity is opposite to the chain atom next to it. The copy comes out as a complement, and copying the complement gives back the original. This is how a catalytic site that has formed once can be copied rather than found again by chance, which turns an astronomically unlikely search into a question of whether the site has ever formed at all.
7. Exchange
A bonded atom can be swapped for a free atom of the same column (one spacer for another spacer, one donor for another donor), keeping its bonds if the incoming atom’s valence and polarity allow them. Exchanges are proposed blindly: any row of the same column, with equal chance, and never across columns. This is the event that lets a distinction stop mattering, and it is the one compression needs.
Nothing runs backward except by breaking or decay. There is no un-copying and no un-transfer.
Where compression comes in
The framework holds that a sovereign structure produces its own order by compression: dropping distinctions that make no difference and keeping only what carries weight. What a lattice lacks names substitution as the one role a bare lattice has no stand-in for, and the one that feeds compression. The grammar is built so that compression can happen during and after these events, with no rule that performs it.
- Exchange proposes. Blind swaps within a column keep testing whether a part could be replaced by a cheaper or more common one.
- Upkeep and decay dispose. A structure pays to hold every bond, and bonds between more reactive atoms decay faster. Where a swap leaves the structure working and cheaper to hold, the cheaper version lasts longer and is copied more often. Where a swap breaks something, the structure loses the part or fails.
- Copying keeps the result. Template copying carries the simplified version forward, so a distinction dropped once stays dropped.
No step chooses the simplification. The grammar only proposes; what survives is decided by the structure’s own costs. Compression is then something the lab reads off a run afterwards: parts that were once specific and are now interchangeable, at lower upkeep, with function kept. The kernel reserves a slot for measuring it, the Compression Descriptor in ALPHA § 8.3, and leaves it undefined. This page does not define it. Before I‑Pop, compression can happen but is not α-trace, since the α-trace exists only for a sovereign structure. After I‑Pop, the pattern those copies carry is what the framework reads as α-trace.
The Chemistry page’s claim that substitution error should grow with row distance can be tested directly: exchanges are proposed at every row distance with equal chance, so any rise in failures with distance comes from the chemistry and not from the proposal rule.
The reactor: the order things happen in
- One atom per cell. A cell holds at most one atom, its bonds, and a count of tokens.
- Choose which cells act. In each step every cell acts with probability u, the substrate’s time dial. At u = 1 every cell acts together, which reproduces the ordinary synchronous update exactly; below 1 the shared clock is removed. Steps 3 to 5 apply to each cell that acts.
- List what it could do. The cell lists every event above that is admissible with its neighbours, counted by Duncan’s Neighborhood, whose cost can be paid from the tokens present.
- Pick one, weighted by reactivity. More reactive atoms are more likely to take part. Reactivity already sets cost on the Chemistry page; here it also sets how often an atom reacts, which gives the row axis a job beyond price.
- Move. Free atoms and tokens swap at random with empty neighbouring cells. Bonded atoms stay put; a molecule moves only by breaking and re-forming.
- Charge upkeep and apply decay, then record every token spent in the ledger.
The first thing it must do is fail
The Chemistry page set this before anything was written: the first run must show a capability disappearing when the role that supplied it is switched off. The ablations, stated in advance:
| Switched off | What should disappear | If it does not |
|---|---|---|
| catalysis (event 3) | boundaries fail to close within the run, or close far more slowly | catalysis is not doing the work it is credited with |
| template copying (event 6) | catalytic sites form rarely and do not spread | something else is copying, and must be found |
| exchange (event 7) | no interchangeable parts appear; compression reads zero | compression is being produced by some other route |
| the energy flow | every structure decays to free atoms | something persists for free, which the kernel forbids |