Stafford

← 2026-09-16 · all briefs · 2026-09-18 →

Stafford Brief — 2026-09-17

⚙️ OPS — read this first, it is about the system rather than the evidence. Stafford did not run 2026-09-11 through 2026-09-16. Six days. The research-evidence scan ran every one of them and today's landed normally at 12:20 UTC, so no evidence was lost — but the judgment layer was dark while six days of evidence accumulated, and this is the first brief written under the plain-language rules you added to the template on 09-11. Two consequences worth knowing: the counterargument ledger had no entries between 09-10 and today (two are added below, one of them backfilled from the scan's own case file), and no sparring note has been issued since 09-07. Nothing is broken that I can see from inside the session; the gap is reported rather than diagnosed.

Today's scan is consumed, not repeated. It led with Kentaro Toyama's 2011 Law of Amplification as prior art on Technology Illusion, escalated the Previous-Shock Archive pattern, and staged a randomized negotiation experiment (Zhu et al.). All of that is in research-evidence/briefs/2026-09-17.md and is not restated here. This brief is about what the scan's find implies and where its recommendation is wrong.

📖 TODAY'S READ

Your second breakpoint was published as a journal article in 1975, and it is more famous than the one the scan found this morning.

Steven Kerr, "On the Folly of Rewarding A, While Hoping for B," Academy of Management Journal, 1975. The title is the mechanism. Kerr's argument is that people work out what actually gets rewarded and then do that, more or less regardless of what leadership says it wants — and his explanation for why organizations keep doing this is the same as yours: reward systems get built around things that are easy to see and count, while the things that actually matter are hard to measure. So organizations pay for what they can measure and hope for what they meant. That is Incentive Fragmentation, named, explained, and in print for fifty years, in a top-tier journal that is standard reading on most management courses.

What Kerr does not do is the whole of your position. He illustrates a folly and explains why it recurs. He gives no test — nothing a leader can run against their own organization to find out whether this is their problem. He has no account of how incentive misalignment interacts with the other four failure patterns, because he is not describing a set of failures; he is describing one. Your self-checks are the thing he lacks.

So what: this changes what you write, not whether you were right. The scan told you to concede Toyama in a paragraph. You need to concede Kerr in the same breath, or the concession looks selective — see CHALLENGE below. Filed at knowledge-base/kerr-folly-rewarding-a-while-hoping-for-b-1975-12.md.

One honest limit, stated up front. I could not read Kerr's actual article. The free copy online is a photograph of the pages with no readable text, and the publisher's site refused the request. Everything above comes from the published abstract and several independent descriptions of the paper. I am confident the antecedent exists and says what I say it says — that is established by the title and abstract alone — but nothing in the file is quoted from inside the article, and I have not tagged any Four Forces on it. Same limit applies to Kotter below.

🔭 THINKER TO WATCH

Steven Kerr — the 1975 article is his; he later ran learning and development at General Electric under Welch and then at Goldman Sachs, which makes him unusually hard to dismiss as an academic who never had to make a payroll.

Why he matters right now rather than as history: he revisited the argument himself in a 2017 piece in Academy of Management Perspectives titled "40 Years (and Counting)." I have the title and the venue and nothing else — I have not read it and am not characterising what it says. The reason to flag it: if Kerr has publicly updated his own 1975 claim, that is either the strongest possible citation for your Antecedents section or the most awkward one, depending on what he concluded, and it is a fifteen-minute read to find out. Worth doing before you draft.

How your position differs, stated exactly, because this is the sentence a hostile reader will test. Kerr tells you that organizations reward the wrong thing and why that keeps happening. He does not tell you whether your organization is doing it, how to find out, or what to do about it — and he does not connect it to anything else that goes wrong at the same time. You supply a test and a set; he supplies an observation. That is a real difference and it is the one to defend. It is also narrower than "we identified five breakpoints," and narrower is the point.

🏢 CASE IN THE WILD

Nothing new from my lane today, and I would rather say that than pad it. The scan's case — the randomized negotiation experiment where the AI was better than the humans in every arrangement and the humans edited its value away — is genuinely good and is in today's scan brief. I have mirrored it into the counterargument ledger because it cuts against your human-oversight arguments and it had not been logged there. I have not restated it here.

The closest thing to a case I have is uncomfortable and it is yours. You already cite Hammer & Champy's Reengineering the Corporation in Paper 3 — in the endnote where you trace where the 70% figure actually came from. That book's explicit subject is structural friction and broken handoffs blocking execution, which is Process Friction. The citation has been sitting in your own footnotes doing arithmetic while the lineage claim went unmade.

💭 OPEN QUESTION

If four of the five breakpoints were already named by somebody else, what exactly is the contribution — and would you still have one if a reader granted you only part of it?

My answer is that there are three candidates. The taxonomy: nobody has put these five together, because they come from five separate literatures in four different fields across four decades, and those literatures do not cite each other. The diagnostic: none of the antecedents gives an organization a test to run on itself, and you do. The speed: none of them has any account of these patterns operating fast — Kotter's own timeline runs to years, with the deepest change arriving around year five.

The question I cannot answer today is whether the first one survives, and it is the one carrying the most weight. Has anyone already published a framework that treats three or more of these five as a single interacting system with a diagnostic attached? I have not run that search. If the answer is no, your contribution claim is strong and the Antecedents section is a victory lap. If the answer is yes and you have not cited it, this stops being a section and becomes a revision. The names I would check first are Beer & Nohria's "Cracking the Code of Change" (2000), Burke–Litwin, the McKinsey 7-S lineage, Pettigrew, and — the one I would actually bet on — CFIR, the implementation-science framework, which is explicitly a multi-domain diagnostic taxonomy. That search is the next exploratory slot.

🥊 CHALLENGE

The strongest available attack on your entire body of work, stated at full strength: Five Breakpoints is change-management canon with new labels on it.

Four of five, and the exposure is worse than it looks, because of which four. Toyama is obscure enough that conceding him reads as scholarship. Kotter is not. "Declaring victory too soon" is a phrase in ordinary executive circulation — nobody has to find it. Any reader who has taken a management course meets Momentum Mirage already knowing it, and meets Incentive Fragmentation already knowing Kerr.

Which is why I think the scan's routing recommendation is wrong, and this is the main thing I want to put in front of you today. Its words: "One paragraph and one citation closes it permanently." It does not. One paragraph conceding the obscure 2011 antecedent, with no mention of Kerr or Kotter, does not read as honesty — it reads as a writer who conceded the antecedent he happened to trip over and missed the two his readers already know. That is worse than saying nothing, because it invites the reader to check whether you looked. The exposure is structural, so the concession has to be structural: one Antecedents section covering all five.

What would make this objection win outright: a prior work that does the integration — three or more of these five treated as one interacting system, with a diagnostic. See OPEN QUESTION; that search decides it. What makes it survivable, and all three are real: nobody integrated these literatures; nobody supplied a self-check; nobody had an account of speed. Logged in full as The relabeling objection, status CONTESTED, in knowledge/thesis-challenges.md.

📈 PATTERN BUILDING

New pattern — The Antecedent Ledger. Four sources, four decades, four fields, established today in one sweep. Kerr (organizational behaviour, 1975), Hammer & Champy (process management, 1993), Kotter (change management, 1995), Mankins & Steele (strategy, 2005), plus Toyama (international development, 2011) from the scan. Five breakpoints, five antecedents, and not one of these five literatures cites any of the others.

That non-citation is the pattern's actual content, and it cuts in your favour. The reason no one framework covers all five is that the five failure patterns were each discovered inside a different discipline, by people who had no reason to read each other. Kerr's readers are organizational behaviour people; Hammer's are process engineers; Toyama's are development researchers. Your framework is the first thing in this base that treats them as one object. That is a defensible contribution and it is a different, smaller claim than the one Paper 2 currently implies.

What would break it: the integration search in OPEN QUESTION. Until that runs, this pattern is evidence that the antecedents are scattered, not proof that nobody has gathered them. Proposed as a new entry in the scan's pattern file — see Routing.

🧱 ASSET SUGGESTION

"What the Five Breakpoints Owe" — the Antecedents section, plus an optional companion post.

The argument in two sentences: name all five antecedents yourself, precisely and without being asked, then state the narrower contribution claim — integration across literatures that do not talk to each other, a diagnostic none of them supplies, and a speed claim none of them could have made. Conceding lineage exactly is the same public move as Paper 3 retiring the 70% figure, and a second instance is what turns that from an incident into a practice you are known for.

Form: a section inside Paper 2, not a standalone essay. The concession has to travel with the claim; a separate essay leaves the paper unguarded. The companion post ("Five ideas I didn't invent, and the one I think I did") is optional and secondary.

This is argument-priced, not time-priced, so it is not a publication window and I have not flagged one. Nothing expires. What makes it urgent anyway is the asymmetry: writing it costs one section, and not writing it costs you the chance to be the one who said it first. Full spec at knowledge/asset-suggestions.md #8. It is blocked on one thing — the integration search — and that should run before you draft, not after.

🎓 SPARRING NOTE

Growth edge #1 has been stated too narrowly since 2026-08-29, and today is the correction.

The edge reads: ground Claim 1 in the formal literature, because skipping it invites a referee-shaped ambush. Every hook parked against it has been about Claim 1's hierarchy-and-information lineage. The ambush is not waiting on Claim 1. It is already loaded, against four of five breakpoints, and springing it requires no referee — it requires one management course.

The prompt, and attempt it before reading Kerr or Kotter, because the point is the decision and not the reading. Write one paragraph, addressed to a hostile reader who knows Kotter cold, stating what Five Breakpoints claims that Kotter's eight errors do not. Then answer what that paragraph invites: is the contribution the taxonomy, the diagnostic, or the speed claim — and can you defend it if a reader grants you only one of the three?

Fifteen minutes, and it cannot be answered by gathering more evidence, which is what the rest of this system is for. Logged at knowledge/brandon-development.md.

SOURCES & THREADS

Exploratory slot — spent on a prior-art sweep, which neither query log has ever run. Not a case-file gap. CLAUDE.md operating rule 2 puts author and instrument verification in my lane, and this was the obvious follow-up question to the scan's own find: if one breakpoint has a canonical antecedent, do the other four? Rule-12 check first: I grepped your papers, the book manuscript and the framework source for Toyama, amplification, Agre, Heeks and "prior art" and got zero hits — so the exposure is genuinely unclosed and this is not handing you your own argument back. Four searches, one per remaining breakpoint, plus a citation-pinning pass. Yield: 4 antecedents confirmed, 2 KB files, 1 challenge, 1 asset, 1 sparring note. Logged with the full reasoning in knowledge/query-log.md.

Primary-read failure, recorded rather than glossed. Kerr and Kotter could not be read. Both open PDFs are image scans or font-subset reprints with no extractable text; the publishers return 403 or a paywall. I installed nothing and worked around nothing — I tried four sources across two papers and stopped. Both KB files carry no Four Forces tags and claim nothing from inside the article bodies. Mankins & Steele and Hammer & Champy are verified bibliographically but deliberately not ingested — I have not read either, and a KB file built on a search summary is exactly the contamination the hard rules exist to prevent.

Threads: none due today. Nearest is the agent-governance thread on 2026-09-30. The scan advanced the team-agent thread to 11-17 and opened a new amplification thread (next check 10-17) — both are its work, not mine, and I have not duplicated them.

Candidates: two logged, both as unread leads only — Kerr's 2017 Academy of Management Perspectives retrospective, and the change-management-canon stream generally (Beer & Nohria, Burke–Litwin, CFIR) as the integration search's target list. No promotions, no expiries.

No window flag. Nothing found today has a clock. The Antecedents work is urgent by asymmetry, not by deadline, and calling that a window would be manufacturing urgency.

Routing

strategy — the one thing to act on, and it corrects this morning's scan

The scan's routing said the Toyama prior-art problem is closed by "one paragraph and one citation." Do not do that. Four of five breakpoints have canonical antecedents and two of them (Kerr, Kotter) are far better known than Toyama. A selective concession is worse than none. Write one Antecedents section covering all five, and run the integration search first — if some existing framework already gathers three or more of these with a diagnostic attached, this is a revision rather than a section. Full argument: knowledge/thesis-challenges.md, The relabeling objection (CONTESTED). Asset spec: knowledge/asset-suggestions.md #8.

personal — one line

The gap that matters in your development file is not scholarship. It is conceding specifically rather than vaguely, early rather than under pressure — and you have already demonstrated you can do it once, with the 70% figure. This is the same move on better material.

STAGE-CANDIDATE — ready to run in research-evidence

Two entries, both with forces: [] deliberately (primaries unreadable — honest nulls, not missing data).

python3 build/add-research-json.py --json '{
  "title": "On the Folly of Rewarding A, While Hoping for B",
  "publisher": "Academy of Management Journal",
  "url": "https://journals.aom.org/doi/abs/10.5465/255378",
  "published_date": "1975-12-01",
  "credibility": "high",
  "breakpoints": [
    {"name": "incentive_fragmentation", "rationale": "Kerr'"'"'s 1975 title thesis — organizations reward A while hoping for B, and people do A — states the Incentive Fragmentation mechanism thirty years before the framework, with the same stated cause: reward structures are built on measurable and visible criteria while the objectives that matter are subjective and hard to measure."}
  ],
  "forces": [],
  "key_findings": ["Prior art on Incentive Fragmentation: the breakpoint mechanism named in Academy of Management Journal 18(4), 1975, pp. 769-783", "Kerr offers no diagnostic instrument and no account of interaction with other failure patterns — illustration, not taxonomy"],
  "instrument": "Illustrative essay; no sample, no measurement, no test. Primary not readable (image scan, no text layer; publisher 403) — entry claims nothing from inside the article body."
}'

python3 build/add-research-json.py --json '{
  "title": "Leading Change: Why Transformation Efforts Fail",
  "publisher": "Harvard Business Review",
  "url": "https://hbr.org/1995/03/leading-change-why-transformation-efforts-fail-2",
  "published_date": "1995-03-01",
  "credibility": "high",
  "breakpoints": [
    {"name": "momentum_mirage", "rationale": "Kotter'"'"'s seventh error — declaring victory too soon, in which visible early progress is read as completion and the effort stops short of the depth that makes it durable — is the Momentum Mirage mechanism named in 1995 in the most widely read article in the change-management literature."},
    {"name": "strategic_disconnection", "rationale": "Kotter'"'"'s third and fourth errors — proceeding without a clear vision, and under-communicating the vision that exists — are the antecedent of vague purpose producing the illusion of alignment, though framed as failed transmission rather than as agreement surviving because the goal is too unspecific to disagree with."}
  ],
  "forces": [],
  "key_findings": ["Prior art on two breakpoints: HBR 73(2), March-April 1995, pp. 59-67, eight errors including declaring victory too soon", "Kotter supplies a prescriptive sequence and no diagnostic — nothing tells a leader which of the eight errors their organization is committing", "No account of speed: Kotter'"'"'s own timeline runs to years, with peak change around year five — the machine-speed half of the thesis is untouched by every antecedent found"],
  "instrument": "Practitioner synthesis of unstructured observation across 100+ companies; no disclosed sample frame, no measurement, no comparison group. Primary not readable (paywall; open copies are image scans / font-subset reprints) — entry claims nothing from inside the article body."
}'

PROPOSED-UPDATE — knowledge/emerging-patterns.md (new pattern, NOT escalated)

The Antecedent Ledger — 4 members, 4 fields, 4 decades. NOT escalated. Claim: each of the Five Breakpoints has a named, dated antecedent in a different literature, and those literatures do not cite one another. Members (2026-09-17, all in one sweep — which is the pattern's own weakness, see below): Incentive Fragmentation ← Kerr, AMJ 18(4), 1975. Process Friction ← Hammer & Champy, Reengineering the Corporation, 1993. Momentum Mirage ← Kotter, HBR 73(2), 1995, error 7. Strategic Disconnection ← Kotter errors 3–4 + Mankins & Steele, HBR, 2005. Technology Illusion ← Toyama, iConference, 2011 (added by the scan the same day). Why it does not escalate, and the rule is being applied against this pipeline's own interest for the fifth consecutive scan: all four non-Toyama members were found in a single query slot by one agent composing searches it expected to succeed. That is one instrument, not four independent observations, and a pattern established by a sweep designed to find it has not yet been tested. Two members (Mankins & Steele, Hammer & Champy) are verified bibliographically only and are not in the base. Two more (Kerr, Kotter) are filed with primaries unread. What would escalate it: an antecedent for any breakpoint surfacing independently, from a scan slot not aimed at prior art. What would falsify the interesting half: a published framework integrating three or more of these five as one interacting system with a diagnostic attached. Check Beer & Nohria (2000), Burke–Litwin, McKinsey 7-S, Pettigrew, and CFIR. Until that search runs, the "nobody integrated them" claim is untested.

PROPOSED-UPDATE — knowledge/paper-2-case-file.md, Counterarguments section (append to the 2026-09-17 addition)

The Toyama concession is one fifth of the exposure, and the recommendation attached to it should be amended. The 2026-09-17 routing note reads "One paragraph and one citation closes it permanently." A Stafford prior-art sweep the same day established antecedents for three further breakpoints: Incentive Fragmentation ← Kerr (AMJ 1975); Momentum Mirage ← Kotter error 7 (HBR 1995); Strategic Disconnection ← Kotter errors 3–4 + Mankins & Steele (HBR 2005); Process Friction ← Hammer & Champy (1993), already cited in Paper 3's 70%-figure endnote. Kerr and Kotter are substantially better known than Toyama, so a concession naming only Toyama reads as incomplete scholarship rather than as honest positioning. Amend the recommendation to a single Antecedents section covering all five, and note that the contribution claim then rests on three legs — integration across non-communicating literatures, a diagnostic none of the antecedents supplies, and the machine-speed extension none of them could have made — of which only the third is currently evidenced by this file.

PROPOSED-UPDATE — knowledge/source-candidates.md

Kerr 2017 retrospective — Academy of Management Perspectives, "40 Years (and Counting): Steve Kerr Reflections on the 'Folly'". First seen 2026-09-17 via the prior-art sweep. 0 signals — title and venue verified, content unread. Chase before drafting the Antecedents section: if Kerr has publicly updated his own 1975 claim, it is either the strongest citation available or the most awkward. Change-management canon as an integration-search target set — Beer & Nohria (HBR 2000), Burke–Litwin, McKinsey 7-S, Pettigrew, CFIR. First seen 2026-09-17. 0 signals. Logged so the next slot reaches the primaries rather than re-deriving the list.

PROPOSED-UPDATE — knowledge/query-log.md (scan's copy), next-slot note

Named but unrun, and the highest-value slot available: the integration search. Has anyone published a framework treating three or more of the Five Breakpoints as one interacting system with a diagnostic attached? This decides whether the relabeling objection (stafford-research/knowledge/thesis-challenges.md, CONTESTED) is survivable or fatal. Check CFIR first.

governance / market — nothing today.