Working Design Document · Not Canon

Eco 2.0 Design: Building a Branch That Ignites

The one change that would open a real bootstrapping interval

Status. This is an engineering design page, not canon and not a lesson. It records the target for the next Bio-Kernel branch and the pass/fail test that decides whether the branch succeeded. It is checked against the working memo “The Bootstrapping Interval Cannot Be Faked — It Can Be Earned.” It introduces no kernel constructs and modifies no canon. Everything here is provisional and expected to change as the design is built and tested.

Eco 1.0 is an attractlet. It fails sovereignty in exactly one place: its boundary is handed to it. Eco 2.0 is the attempt to remove that one piece of scaffolding — to build a system that produces and defends its own boundary — and so to make a real bootstrapping interval open. We do not paint the interval on. We try to earn it, and let the metrics tell us whether we did.

Where Eco 1.0 stands against the four conditions

The design target follows directly from a clean audit of Eco 1.0 against the four sovereignty conditions. Three are met or nearly met; one is not, and it is the decisive one.

Condition 1 · Recursion LockPartial. Feedback runs at the population and field level — density shapes forage, forage shapes energy, energy shapes the next population. But the loop rides on external scaffolding rather than closing purely on its own. Self-reference is present; the lock is not clean.
Condition 2 · Internal Recurcline PersistenceNot yet applicable. This condition presupposes recursion lock. Until the lock closes, there is no sovereign recursion whose identity could persist, and no recurcline to store. It cannot be assessed before Condition 1 and Condition 3 are met.
Condition 4 · Maintenance-Bearing ContinuationMet. Agents pay a metabolic cost every tik and forage the energy to cover it themselves. There is no free-persistence regime. This part is genuinely endogenous already — the bill is real and the agents pay it from their own work.
Condition 3 · Boundary RetentionThe gap. Borrowed. The kernel is exact here: the identity-bearing boundary must be self-produced and self-maintained by the structure's own activity. Eco 1.0's boundary is the refuge terrain, generated once at startup and then fixed. It is handed in, not made. This is the decisive failure — without a self-produced boundary a configuration is at most a coordination structure, never sovereign. It is also the one condition whose repair would let the others fall into place.
The pivot. The kernel names a specific mechanism for this condition: boundary-stabilizing autocatalysis. A configuration that does not self-produce its own identity-bearing boundary cannot satisfy Condition 3, and therefore cannot be sovereign. This is the pivot on which the whole classification turns. Eco 2.0's entire design is aimed at this one mechanism.

The design target

Replace the fixed refuge with a boundary the system makes and holds itself. In Eco 1.0 the refuge is a terrain flag set by the initializer — a wall painted on the map before anything lives. In Eco 2.0 the protective structure must be produced by the agents' own activity and dissolve if that activity stops. The wall must be something the system is continuously building, not something it was given. If the life stops working, the boundary must go with it — because a boundary that outlives the activity that made it is a handed-in wall wearing a disguise.

This single change is chosen deliberately over the alternatives. It is the condition Eco 1.0 fails, it is the pivot the kernel identifies, and repairing it is the most direct route from attractlet to sovereign. The other scaffolding — the external seasonal forcing — is secondary and can be addressed in a later branch once the boundary is endogenous.

Candidate mechanisms (none endorsed yet)

What does a self-produced, self-defended boundary look like in a 100×100 cellular-automaton substrate? These are directions to prototype and test, not decisions. Each ties the boundary to the system's own work so that it cannot outlive that work.

A. Built structure the population maintains. A distinct cell state — call it a wall or shell — that agents actively deposit at a metabolic cost, that decays on its own over time, and that predators cannot cross. The refuge then exists only where the herbivores are currently paying to keep it up. Stop paying, and it rots away. The boundary becomes a running expenditure rather than a fixed feature.
B. Boundary coupled to maintenance (Conditions 3 and 4 fused). Make the wall's persistence draw on the same energy budget the agents spend to live. Then Condition 3 and Condition 4 become one mechanism: the structure pays to hold its own edge, and the edge dissolves precisely when the maintenance stops. This is the tightest fit to the kernel's own coupling of boundary and cost.
C. Interior held different by the interior's own work. Rather than a wall, a maintained gradient the interior actively sustains — a compartment kept in a state genuinely unlike the outside because the agents inside are continuously doing the work to keep it so. The caution here is real: the kernel warns that a mere gradient is not a boundary. This option must produce a defended partition, not just a slope, or it fails Condition 3 for the same reason a flame does.
Failure to watch for. Any of these can quietly reproduce the attractlet. If the “self-made” boundary is really seeded once and then merely maintained on autopilot by a rule the initializer wrote, it is still handed in. The test is whether the boundary is generated and re-generated by the agents' own activity, tik after tik, and whether it fails when that activity fails. If it can outlive the life inside it, it is scaffolding.

The pass/fail test

This is the part that keeps the whole effort honest, and it is the reason the design page exists. Eco 2.0 does not get to declare victory. There is one test, and it is not aesthetic.

Does a recurcline become definable across an interval? Ask the new system for its recurcline. For Eco 1.0 the honest answer is undefined — there is no recursion lock storing it. That answer is the correct diagnosis of an attractlet. Build Eco 2.0 and ask again. If the answer is still “undefined,” the branch is still an attractlet and the boundary is still borrowed. The day the answer changes to a real, accumulating quantity across a span bounded by ignition and loss — that change is the proof of ignition. A real bootstrapping interval has opened, and the system has left attractlet status.

The metrics are not decoration we add after we succeed. They are the instrument that detects whether we succeeded. The switch from “recurcline undefined” to “recurcline defined” is the alarm bell that says: this one is no longer a mimic. And the guardrail runs the other way too — the claim “Eco 2.0 has recurcline” is identical to the claim “Eco 2.0 is a sovereign attractor,” and inherits the full burden of the four-condition test. You cannot produce the readout without building the real thing. That is exactly what makes it a trustworthy test.

What success would look like, concretely

If Eco 2.0 works, the sequence to watch for is:

1. Agents begin depositing and maintaining a boundary out of their own metabolic budget.
2. The protected interior begins feeding the process that maintains the protection — the loop closes on itself, with no fixed terrain doing the work. Recursion lock (Condition 1) becomes clean.
3. The closed loop begins storing a compressed record of what has survived — recurcline becomes definable. Condition 2 comes online.
4. The boundary is now self-produced and self-defended (Condition 3), paid for continuously (Condition 4). All four hold.
5. A bootstrapping interval is now open. It began at the moment the loop locked. It will end if maintenance fails past recovery — at which point the recurcline vanishes and only inert residue remains.

That interval — born at step 2–3, closed at eventual loss — is the artificial bootstrapping interval, in the only sense the phrase is ever allowed to mean: a real interval in an engineered system. Not a faked one. An earned one.

Next steps

1. Prototype mechanism B (boundary fused to maintenance) first — it is the tightest fit to the kernel's coupling of Condition 3 and Condition 4, and the least likely to smuggle in a handed-in wall. Done: a first prototype is running on the live Eco 2.0 page.
2. Port to the live grid so the boundary can be watched forming and dissolving, the way the Eco 1.0 live view shows populations. Done in that prototype — walls appear where herbivores pay to raise them and fade where upkeep stops.
3. Instrument the recurcline-definable test as the explicit success criterion, not an afterthought. Done: the prototype runs an ignition test built to be able to report “not igniting,” which is the honest default for an early build.
4. If it ignites, the bootstrapping-interval metrics become teachable on a real positive example for the first time — which is the companion teaching page the memo calls for next. Gate: that page is unlocked only when the live prototype's ignition test turns green and holds — not before. Until then the metrics have no real positive example, and the teaching page would have nothing honest to stand on.

This is a working design document for a downstream Bio-Kernel experiment. It applies the canonical framework and is judged by it; it does not modify it. The four sovereignty conditions and the boundary-stabilizing-autocatalysis pivot are stated per the kernel's Tier-Sov specification; the recurcline domain-of-definition test is stated per the kernel's Tier-3 specification. All mechanisms are candidate directions, none endorsed, subject to the boundary notice governing the Bio-Kernel Series and to the working memo this page is checked against.

← Back to the Eco 1.0 mapping