The kT-bit Catalog

The first three settle whether it’s a kT-bit at all; the fourth tells you where to look. It’s a working filter, held loosely — when the first three land squarely the fit is good, and when one wobbles, that’s what the confidence score is for.

1. Is free energy dissipating through it? Call that dissipation the flow. Something moves through the structure — water, electrons, blood, sap, air, magma, ions, or a token that gates access to a free-energy flow, like money — and its flow rate is, directly or indirectly, a measure of dissipation. This also reads state: flow cut and nothing dissipating means dead, though still a kT-bit if dissipation through competing paths once built it.
2. Does that flow have more than one adaptive path to take? The pathways are adaptive — the structure explores, branches, and locks on as the flow runs through it (“the container will adapt to the flow”). This is broader than memristive memory: a vortex selecting one spin by positive feedback is adaptive exactly as much as an eroding channel is. A rigid, unchanging pipe is not a kT-bit; an adaptive container is.
3. Do those paths compete for the particle that carries the flow? The bit is the competition — the fork, or the competing spins — never a single channel. A single conduction pathway only carries the flow; it takes two or more to choose. A lone synapse, a lone memristor, a single river reach is half a bit. The pair is the bit.
4. Which way does the flow run — outward to a fork, or inward to a merge? Outward, the fork is in front of you and the bit evaluates. Inward, the junctions just merge, so the choosing fork has climbed a level up — to the divide, not the confluence.

Three yeses and you’re looking at a living, non-collapsed kT-bit. Lose the first and it’s dead. Keep the first but lose the third and it’s collapsed — energy still pours through, but down a single path that already won.

• Fork / evaluate — flow runs outward, one-to-many; a particle picks a path; exploration.
• Merge / harvest — flow runs inward, many-to-one; a scattered resource gets gathered; the choosing fork has climbed one level up, from the confluence to the ridgeline.
• Hebbian (lock-on) — winner-take-flow positive feedback collapses the bit onto one path.
• Anti-Hebbian (reset) — the win undermines its own advantage, the mixture comes back, the search reopens.

Lock-on (Hebbian) — how a path wins and keeps winning:

Reset (anti-Hebbian) — how the win comes undone:

List the conduction resources that flow, and check each has a genuine representative. Stress, strain, surface energy, landscape-position — and matter that simply deposits onto a growing crystal — are excluded; none of them is a flow running through a conduit.

Confidence should be quantitative, not a judgment call. It is the sum of three verdicts, one for each gating question:

Confidence = Q1 + Q2 + Q3

Q4 (fork or merge) is orientation, not a gate, and never scores. Reversibility — whether a bit has an anti-Hebbian reset — is not a gate either; a permanently collapsed bit is still a bit. Every entry is scored in its alive, operating regime.

Each gating question takes one of three values.

Q1 — Is free energy dissipating through it?

• 2 — a particle flows that carries or gates free-energy dissipation, and its flow rate is, directly or indirectly, a measure of that dissipation. A token that gates access to a dissipative flow, like money, scores 2, identical to a physical carrier (water, charge, blood).
• 1 — what flows is a signal that triggers dissipation elsewhere, not the dissipative flow itself; flow rate is not a clean measure of dissipation.
• 0 — nothing flows, or the flow is metaphorical. → cut list.

Q2 — Is the container adaptive? Tested by perturbation, not by motion: block or divert the flow and watch.

• 2 — the structure re-routes, regrows, or re-balances (a dammed river cuts a new channel; an occluded artery grows collaterals; a plugged karst conduit dissolves another). Reaching steady state does not lower the score — latent capacity to remodel under perturbation is what counts.
• 1 — the adaptive response is weak or nearly absent.
• 0 — perturbation does nothing; a rigid pipe. → cut list.

Q3 — Do the paths compete for the particle?

• 2 — a clear set of competing pathways or spins. Direction is irrelevant: a merge/harvest network scores the same as a fork, the competition having only climbed to the divide.
• 1 — the second competing pathway takes an argument to identify (a single forming filament with no obvious rival; a lone radial axis). Competing spins do not sit here — clockwise versus counter-clockwise is a clear, mutually exclusive pair and scores 2.
• 0 — one channel, no rival. → cut list.

A 0 on any gate disqualifies the candidate and sends it to the cut list, so every catalogued entry scores 3–6:

Each table carries the three verdicts in its own Q1, Q2, Q3 columns, and Conf records their sum.