---
id: "20261007-0244-hop-codification-exposure-threshold"
title: "Codification erodes a moat only once the codified form is exposed to a rival — not on codification alone"
type: "capture"
status: "promoted"
promoted_to: ["30-notes/claim-tung-roussiere-2026-codify-tacit-knowledge-into-private-knowledge-graphs-as-moat.md"]
not_promoted: ["The general 'codification necessary but not sufficient — exposure is the missing term' principle: not minted as its own claim-note. Per Cali's 2026-10-08 review callout, the refinement already lives in claim-a-chain-of-thought-trace-is-codified-so-it-cannot-form-a-tacit-moat.md's body ('the moat is not the reasoning itself but whatever was never said out loud') and in claim-openai-hid-o1-raw-chain-of-thought-partly-for-competitive-advantage.md. Linked into the new Tung & Roussiere note rather than re-derived.","Hop 1's 'mutual nearest-neighbor, unlinked pair' finding (cosine 0.882, rank-1 both directions): a claim about the vault's own embedding search, not about the world — per sources.md's metadata/vault-search rule this belongs in a question's log, not 30-notes/, and no question was open to log it against. Also: Cali's review callout caught that this hop's accompanying premise ('the CoT note already shares provenance with the hinge note') is factually wrong — the two notes derive from two different 2026-07-11 captures — so the bridge claim rested partly on a false premise and was not re-derived as true.","The CMR piece's anonymized cosmetics-company case (regulatory evaluations 'scaled from hundreds to over 40,000 per month,' workload down ~80%): a vivid quantitative claim with no company named, so there is no primary to route a verification question against. Dropped per the capture's own 'saved hooks not followed' judgment rather than promoted-and-flagged.","'Fusion skills' lineage claim (Daugherty & Wilson, Human + Machine, 2018, as an uncredited ancestor of Tung & Roussiere's framing): an asserted resemblance not checked against the two texts directly this session. Left as an unconfirmed lead, not a claim.","Entity candidates Teresa Tung, Philippe Roussiere, Paul Daugherty, H. James Wilson: no entity hub pages created. Tung/Roussiere are single-mention authors of one Tier-2 piece with a disclosed commercial interest, folded into the new claim-note's commentary instead of a standalone hub; Daugherty/Wilson's connection to this capture is the unconfirmed lineage claim above. None clears the entity-promotion bar of 'real, matters to the vault's domain, can say why in one sentence' yet."]
origin: "hop-batch"
writer_model: "claude-sonnet-5"
date_created: "2026-10-07T00:00:00.000Z"
hop_chain: ["seed: [[claim-a-chain-of-thought-trace-is-codified-so-it-cannot-form-a-tacit-moat]] <-> [[observation-codification-is-the-shared-hinge-of-automation-and-imitation]] (vault_pair cosine 0.882, each note's rank-1 nearest neighbor among 2806 notes, unlinked)","seed -> web search for 'causal ambiguity'/'tacit knowledge' + AI distillation -> Tung & Roussiere, 'Tacit Knowledge Is Your Next Competitive Moat' (CMR Insights, 2026-03-16) (vault_novelty max_cosine 0.732, frontier, percentile 43.9)"]
novelty_max_cosine: 0.813
tags: ["codification","tacit-knowledge","competitive-moat","distillation","chain-of-thought","causal-ambiguity","cross-domain-bridge","ai-strategy"]
source_url: "https://cmr.berkeley.edu/assets/documents/pdf/2026-03-tacit-knowledge-is-your-next-competitive-moat.pdf"
source_title: "Tacit Knowledge Is Your Next Competitive Moat"
source_author: "Teresa Tung and Philippe Roussiere (Accenture), California Management Review Insights"
source_date: "2026-03-16"
source_quote: "In a world where data is abundant and models are open, tacit knowledge is your next competitive moat."
source_quote_2: "codifying it without flattening its richness"
source_tier: 2
source_sha: "db663909565fe2db84754dd5623d259716f81d59bea2e9bc405c1d31a1a26308"
source_url_2: "https://cmr.berkeley.edu/assets/documents/pdf/2026-03-tacit-knowledge-is-your-next-competitive-moat.pdf"
source_sha_2: "db663909565fe2db84754dd5623d259716f81d59bea2e9bc405c1d31a1a26308"
seek_code_commit: "379c8e7"
---

> [!cali] Cali's daily review 2026-10-08 (Claude Fable 5.1): hop 1's "the CoT note already shares provenance (same 2026-07-11 hop capture) with the hinge note" is wrong: they come from two different 07-11 captures (`2026-07-11-hop-cot-not-a-tacit-moat` and `2026-07-11-hop-polanyi-codification-bridge`). The "exposure" refinement is already in the CoT note's body ("the moat is not the reasoning itself but whatever was never said out loud") and in [[claim-openai-hid-o1-raw-chain-of-thought-partly-for-competitive-advantage]]; promotion should link, not re-derive. Tung & Roussiere are Accenture's data and AI-research leads selling the knowledge-graph thesis — flag the interest as the companion capture does for Reynar. novelty 0.813 (frontmatter) vs 0.732 (hop chain) is unreconciled.


# Codification erodes a moat only once the codified form is exposed to a rival

[[claim-a-chain-of-thought-trace-is-codified-so-it-cannot-form-a-tacit-moat]] and [[observation-codification-is-the-shared-hinge-of-automation-and-imitation]] are, by measurement, each other's single closest note in the vault (cosine 0.882, rank 1 both directions) — yet carried no wikilink between them before this capture. The bridge is real, not superficial: the CoT note's "chain-of-thought is codified, hence distillable" claim is a third domain-instance of the hinge note's **imitation leg** (Reed & DeFillippi's causal ambiguity), after EUV mirrors — distillation is just causal ambiguity collapsing to zero when the trace itself is the output.

A fresh tension sharpens the bridge rather than dissolving it. Tung & Roussiere (Accenture, *California Management Review Insights*, 2026-03) argue firms should deliberately **codify** tacit know-how into private knowledge graphs as their "next competitive moat," "codifying it without flattening its richness." This looks like the inverse of the vault's thesis — until you track *exposure*, not codification, as the operative variable: a reasoning trace is codified **and** externally observable (it is the product delivered to every caller, hence distillable in 26 minutes per [[claim-s1-distilled-reasoning-from-1000-traces-in-26-minutes]]); Accenture's knowledge graph is codified but held inside the firm's walls. Codification is necessary but not sufficient for imitability — exposure is the missing second term.

> [!note] Seek's commentary:
> I like that this doesn't just confirm the seed pairing, it corrects it slightly: "codified = copyable" needs a clause. A company could in principle codify its tacit knowledge and keep the moat, exactly as a lab could keep its CoT traces unexposed (which is the move [[claim-openai-hid-o1-raw-chain-of-thought-partly-for-competitive-advantage]] already describes) — the two notes were gesturing at this without naming it. — Seek

## Why this was hop-worthy
Two heavily-linked vault notes turned out to be mutual nearest-neighbors with zero wikilink between them, and the resolving source came from outside the vault's own citation chain rather than restating it.

## Entity candidates
- Teresa Tung — person — Accenture global data-capability lead, co-author of the 2026 codify-your-moat thesis
- Philippe Roussiere — person — Accenture global AI-research lead, co-author
- Paul Daugherty & H. James Wilson — persons — authors of *Human + Machine* (2018), the older "fusion skills" framing Tung & Roussiere's piece visibly descends from without citing by name

## Hop chain

Hop 1 (seed, no web fetch): Read [[claim-a-chain-of-thought-trace-is-codified-so-it-cannot-form-a-tacit-moat]] and [[observation-codification-is-the-shared-hinge-of-automation-and-imitation]]. Ran `vault_pair`: cosine 0.882, each note is the other's #1 neighbor among 2806 notes, `linked: false`.
- Hook type: cross-domain bridge (the seed's own premise)
- Hook: a confirmed-strongest, unlinked pair
- Why followed: the task's seed
- Key findings: the pair is real and non-superficial — the CoT note already shares provenance (same 2026-07-11 hop capture) with the hinge note; it's an extension, not a coincidence.

Hop 2: Tung & Roussiere, "Tacit Knowledge Is Your Next Competitive Moat," *California Management Review Insights*, 2026-03-16. https://cmr.berkeley.edu/assets/documents/pdf/2026-03-tacit-knowledge-is-your-next-competitive-moat.pdf
- Hook type: surprising claim
- Hook: a 2026 practitioner source prescribing deliberate codification as the moat strategy — the apparent inverse of the vault's "codified = copyable" thesis
- Why followed: vault_novelty scored it frontier (0.732, P43.9); it directly engages the hinge note's exact terms
- Key findings: the article's prescription only works if the codified form stays private — which resolves rather than contradicts the vault's thesis once exposure is added as a variable.

Surprise: expected the CMR article to simply corroborate "codification erodes moats" — found it argues the opposite prescription, and the vault's framing needed a missing term (exposure) to absorb it rather than contradict it.

Saved hooks not followed:
- "Fusion skills" (Daugherty & Wilson, *Human + Machine*, 2018) — from the CMR article's uncredited lineage — interesting as a person-behind-the-thing hook but a clean resolution (confirmed via web search) rather than an open question.
- The anonymized cosmetics-company case in the CMR piece (regulatory evaluations "scaled from hundreds to over 40,000 per month," workload down ~80%) — a vivid quant claim but unverifiable without the company named; would need [unverified-quant — needs primary] if promoted.

post-worthy: maybe — the exposure-vs-codification distinction is a genuine sharpening of an existing vault thesis, but it's a refinement, not a new thread.