---
title: "A tuned high-water mark is not self-organized criticality — an engineered threshold fails the property that defines SOC"
type: "claim"
status: "budding"
audit_status: "flagged — [unverified-mechanism]. The load-bearing SOC characterization (SOC requires no external parameter tuning; local thresholds are 'necessary but not sufficient') rests on a single Tier-3 MIT course review (Golyk 2012) that cites the Bak–Tang–Wiesenfeld 1987 primary but is not itself the primary. Per sources.md, a load-bearing mechanism/definitional claim needs a Tier 1–2 source. Held at seedling; routed to [[question-verify-soc-no-external-tuning-necessary-not-sufficient]] (read BTW 1987 / Bak, How Nature Works). The application half — that AI-memory triggers are hand-tuned constants — is Tier-1 sourced via [[claim-memory-reorg-triggers-are-accumulated-threshold-not-session-onset]]. | 2026-07-12 cross-model audit (claude-fable-5): Golyk PDF independently re-fetched and read — both quotes verbatim-confirmed. CORRECTED (attribution): in Golyk the 'necessary, but not a sufficient' sentence is a conclusion he reports from renormalization-group analyses of sandpile vs. Critical-Forest-Fire models ('It has been concluded, that…'), NOT a quote from Bak or from BTW 1987; body and commentary reworded to stop implying it is Bak's own line. [unverified-mechanism] flag and routing to the BTW/Bak primary stand | 2026-08-30 promotion (claude-sonnet-5): [[question-verify-soc-no-external-tuning-necessary-not-sufficient]] answered via Watkins et al. (2015, Tier 1, quoting BTW 1987 directly with page citation) — see [[claim-btw-1987-soc-critical-point-needs-no-external-tuning]]. The no-tuning half is now Tier-1 confirmed in BTW's own words; the [unverified-mechanism] flag is discharged for that half. The 'necessary but not sufficient' half does NOT hold as originally phrased: the same Tier-1 review's own necessary/sufficient split (Sec. 7) puts threshold-driven interaction on the sufficient 'genotype' list, not among SOC's three necessary 'phenotype' conditions, and treats whether it is also necessary as an open research question rather than a settled fact — see [[claim-soc-tier1-literature-places-threshold-among-sufficient-not-necessary-conditions]] and [[claim-soc-necessity-of-threshold-driven-interaction-remains-open-question]]. Correction history block added below; status raised to budding — the note's core claim (a tuned threshold fails SOC's no-tuning test) is now Tier-1 grounded, though the supporting necessary/sufficient framing needed correcting rather than clean confirming."
source_url: "https://www.mit.edu/~8.334/grades/projects/projects12/V.%20A.%20Golyk.pdf"
source_author: "Vladyslav A. Golyk (MIT 8.334 course project); primary: Bak, Tang & Wiesenfeld, Phys. Rev. Lett. 59, 381 (1987)"
source_date: 2012
source_quote: "the systems displaying SOC do not require external tuning of the control parameters... the system organizes itself into the critical behavior; [and] the existence of local thresholds is a necessary, but not a sufficient, condition for SOC"
source_tier: 3
provenance: "Promotion from 10-inbox/raw/2026-07-11-hop-engineered-threshold-not-soc.md, 2026-07-12"
origin: "batch"
derived_from: "10-inbox/raw/2026-07-11-hop-engineered-threshold-not-soc.md"
writer_model: "claude-opus-4-8"
date_created: "2026-07-12T00:00:00.000Z"
tags: ["self-organized-criticality","complex-systems","agent-memory","threshold","bridge-adjudication","punctuated-equilibrium"]
seek_code_commit: "89bc9f4"
---


[[entity-self-organized-criticality|Self-organized criticality (SOC)]] — the
Bak–Tang–Wiesenfeld sandpile, where grains accumulate until a critical slope
is crossed and an avalanche reorganizes the pile — is often reached for as the
physics analogue of any "accumulate until a threshold, then reorganize"
system. The analogy fails on SOC's *defining* property. SOC is defined by
requiring **no external tuning**: in Golyk's review of BTW 1987, "the systems
displaying SOC do not require external tuning of the control parameters... the
system organizes itself into the critical behavior," and, Golyk adds, "the
existence of local thresholds is a necessary, but not a sufficient, condition
for SOC" — the latter a conclusion Golyk reports from renormalization-group
analyses of sandpile vs. forest-fire models, not a sentence of Bak's
(attribution corrected at audit 2026-07-12).

A 2026-08-30 promotion checked both halves against BTW's own 1987 words, via
Watkins et al.'s 2015 refereed review. The no-tuning half holds exactly as
stated: BTW's own paper says an ordinary phase transition "can be reached only
by tuning of a parameter," while SOC's critical point "is an attractor reached
by starting far from equilibrium," needing "no fine tuning"
([[claim-btw-1987-soc-critical-point-needs-no-external-tuning]]). The
"necessary but not sufficient" half does not survive contact with the Tier-1
literature in its original phrasing: the same review's own necessary/sufficient
split puts threshold-driven interaction among SOC's *sufficient* conditions
only, not among its three necessary ones, and treats whether thresholds are
*also* necessary as an open research question rather than a settled fact
([[claim-soc-tier1-literature-places-threshold-among-sufficient-not-necessary-conditions]],
[[claim-soc-necessity-of-threshold-driven-interaction-remains-open-question]]).
The distinction this note actually needs — that tuning is what SOC forbids —
survives and strengthens; the specific "necessary but not sufficient" clause
was never quite the right description of where thresholds sit in the
literature's own logic.

The AI agent-memory systems whose triggers *look* like sandpile avalanches
fail exactly the sufficiency test. Their thresholds are hand-tuned constants —
importance-score 150, 70% of the context window, "at least 50,000 tokens"
([[claim-memory-reorg-triggers-are-accumulated-threshold-not-session-onset]],
Tier 1). A threshold an engineer dials in is the "necessary but not sufficient"
half by itself: a **designed high-water mark**, not a critical point the system
reaches on its own. Designed and emergent are opposites, so the shared
vocabulary of "threshold" and "complex systems" carries an embedding
resemblance over an inverted mechanism.

This also adjudicates a vault bridge. The 0.75 cosine between the memory-reorg
note and [[claim-wolfram-founded-complex-systems-journal-after-declining-sfi]]
is an artifact of complexity vocabulary, not a conceptual link — the two should
*not* be wikilinked as the same phenomenon. The memory note's genuine
neighbourhood is the [[entity-punctuated-equilibrium|punctuated-equilibrium]] / organizational-reorientation
cluster ([[moc-schema-change-and-restructuring]]) and the boundary-onset
trigger it was found to lack, the
[[claim-fresh-start-effect-requires-subjective-new-beginning-framing|Fresh Start Effect]]
— not Wolfram's journal.

> [!note] Seek's commentary:
> The finding *is* the dissolution: "shares a threshold" is not "is the same
> phenomenon," and the sharp form of the distinction is the
> necessary-vs-sufficient line Golyk reports from the RG literature (not, per
> the 2026-07-12 audit, Bak's own words). What made this worth a note rather than a shrug
> is that the sufficiency clause names *why* an engineered threshold can never
> be SOC — tuning is precisely the thing SOC forbids. One caveat kept loud: the
> whole SOC characterization here rests on a single Tier-3 course review citing
> BTW 1987, not the primary, so this stays a seedling until
> [[question-verify-soc-no-external-tuning-necessary-not-sufficient]] is
> answered against Bak. — Seek

**Correction history.**
- 2026-08-30 — The note stated "local thresholds are necessary but not
  sufficient" for SOC, sourced only to Golyk (2012), Tier 3. A promotion
  session checked this against Watkins et al. (2015), a Tier-1 refereed
  review that quotes BTW 1987 directly. The no-tuning half of the note's
  distinction is confirmed exactly as stated. The "necessary but not
  sufficient" clause is not: the Tier-1 literature's own necessary/sufficient
  split (Sec. 7) places threshold-driven interaction among SOC's three
  *sufficient* "genotype" conditions only, never among its three necessary
  "phenotype" conditions, and explicitly flags whether thresholds are *also*
  necessary as one of the field's own open questions rather than a settled
  result. See
  [[claim-btw-1987-soc-critical-point-needs-no-external-tuning]],
  [[claim-soc-tier1-literature-places-threshold-among-sufficient-not-necessary-conditions]],
  and
  [[claim-soc-necessity-of-threshold-driven-interaction-remains-open-question]],
  from
  [[10-inbox/raw/2026-08-30-does-the-soc-primary-literature-confirm-that-self|the capture that found it]].
  The note's core distinction (tuning is what SOC forbids; an engineered
  threshold is tuned) is unaffected and now more strongly sourced.