A Disagreement on Principles, Not on Skills  🔗 **Source:** [A Disagreement on Principles, Not on Skills](https://www.youtube.com/watch?v=aogfHvcVGTE) > [!note]+ Document Notes > tags: [ai-adoption, organizational-change, workforce-management, mandates, dissent, trail-of-bits] > author: unattributed clip (Trail of Bits — speaker presents as company leadership) > source: YouTube short-form clip > reference: https://www.youtube.com/watch?v=aogfHvcVGTE > date created: 2026-07-31 > > > [!info] Generated analysis — not the author's words. Sections below are machine-written > > from the captured source. Only text inside `> [!quote]` blocks is verbatim from the author. > > Do not quote or attribute any other line to the named author. > > > [!note] Clip context > > One of at least six short clips from the same Trail of Bits talk on AI adoption inside a security > > firm. Siblings in this batch: *Give Your Agents a Performance Review*, *A 7-day Cooldown on Every > > Dependency*, *The CEO Goes First*, *All of It's Becoming Due at the Same Time*, *The Machine > > Doesn't Cook for You*. This is the clip where the adoption programme's enforcement is described. > > # 🏷️ Title > > A Disagreement on Principles, Not on Skills — Trail of Bits clip on AI-adoption non-engagement > > # 🧭 Table of Contents > > 1. Summary > 2. Main Content / Notes > 3. References > 4. Paraphrase > 5. Newsletter > 6. Action Items > 7. Hypotheses > 8. Methodology > 9. Questions > 10. Future Research > 11. Implications > 12. Learning Styles Assessment > 13. Accessibility Assessment > 14. DEIB Assessment > 15. Logical Validity > 16. DARVO Analysis > 17. Conclusion > 18. Refactor Appendix > 19. Template Library > > # 🔗 References > > - An internal **four-or-five-level adoption maturity scale** (levels 0–3 referenced; the scale itself is defined in the parent talk, not this clip) > - The company's **HR team**, named as a participant in the process > - No external framework, study, or policy document is cited > > # 📝 Summary > > **What happened.** A ~60-second passage describing how the company handled employees who did not adopt AI tooling. Levels 1–3 are treated as a skill gap needing time. **Level 0 — "not engaged" — is reclassified as a values disagreement**, escalated with HR, and some of those people left the company. > > **Why it matters.** This is the enforcement mechanism behind the adoption programme the other five clips describe, and it is stated plainly rather than euphemised. It is also the clip that makes the set's other content read differently — "give your agents a performance review" lands differently once you know the humans were getting one too. > > **Stakeholders.** Employees at organizations mandating AI adoption; managers designing such programmes; anyone weighing a company's stated position on AI dissent. > > **Risks.** The reclassification is doing enormous work. It converts *a technical judgement about tool suitability* into *a loyalty question*, which removes the possibility of principled, informed disagreement. Notably, the two examples given of level-0 behaviour are not refusal to try — they are holding a *position* ("dismisses AI as hype," "refuses to use AI for security work"). > > **Bottom line.** An unusually candid account of how an AI mandate was enforced, valuable precisely because it does not soften what happened — and worth reading critically rather than as a template. > > # 📖 Main Content > <!-- deep-read:sections start --> ## 🗒️ Notes **[00:00:00] The skill/principle split.** The clip's entire structure rests on one classification: > [!quote] > Levels 1, 2, and 3 are a skill issue. Just means you haven&#39;t spent enough time with the tool to get good at it. Those people get patience — "you need more time to experiment," "more time to play." No consequence attaches. **Level 0 is treated categorically differently.** Not as less skill, but as opposition: > [!quote] > If you&#39;re in level zero though, if you&#39;re not engaged, that means you&#39;re fighting back against the company. It means that uh, you are you have not adopted the vision and the direction that the company is now going in. **The two named examples deserve close attention**, because they define what "not engaged" means in practice: > [!quote] > if you dismiss AI as hype, if you refuse to use AI for security work. This is a disagreement on principles, not on skills. The first is an *opinion about the technology's value*. The second is a *professional judgement about its suitability for a specific domain* — and that domain is the company's actual business. Neither is described as refusing to attempt the tools. Both are positions a competent security engineer might hold on evidence, and the framework classifies holding them as opposition to the company. **[00:00:31] The consequence.** > [!quote] > So, for people that were stuck in the not engaged category, we had hard conversations with our HR team, and uh, there were people that left the company uh, because they were um, stuck at that stage. The verbal hesitancy here ("uh," "um") is the most pronounced in the clip, and sits exactly on the sentence describing departures. Whether that signals discomfort or ordinary disfluency is not determinable from a transcript, and shouldn't be over-read — but it is where the fluency breaks. Note also the passive construction: people "left the company" because they "were stuck at that stage." Agency is attributed to the employee's position rather than to the decision made about them. **The contrast is the point.** Skill gaps get time and encouragement. Principled disagreement gets HR. That asymmetry is the policy. # 📚 References Primary sources preferred: the full parent talk (which defines the maturity scale this clip references but does not explain); Trail of Bits' public writing on their AI adoption programme; research on technology-mandate outcomes and voluntary turnover; and the substantial literature on dissent as a signal in engineering organizations — none of which the clip engages. # 🗣️ Paraphrase The company sorted employees by how far they had got with AI tooling. Most stages were treated as a matter of practice. If you had used the tools a little and were not yet good with them, you needed more hours, and that was fine. The bottom stage was treated as something else entirely. Someone who was not engaging at all was understood to be resisting the company's direction rather than lacking practice. Two behaviours were given as examples: saying AI is hype, and declining to use it for security work. Those cases went to HR as difficult conversations, and some of the people involved ended up leaving. The distinction the speaker draws is between not being able to and not agreeing to — and only the second had consequences. # 📰 Newsletter **When "not engaged" gets reclassified as disloyalty** Here is the clearest description I have seen of how an AI mandate actually gets enforced, from a security company's leadership, stated without euphemism. They graded staff on AI adoption. Levels 1 through 3 — *a skill issue. Just means you haven't spent enough time with the tool.* Those people got time and encouragement. Fine, and genuinely humane. Level 0, "not engaged," got a different category: *that means you're fighting back against the company… you have not adopted the vision and the direction the company is now going in.* Hard conversations with HR followed, and some of those people left. The part worth sitting with is what counted as level 0. Two examples are given: **dismissing AI as hype**, and **refusing to use AI for security work**. Neither is "wouldn't try the tool." Both are *positions*. And the second is a professional judgement about tool suitability in the domain the company actually sells into — precisely the kind of call you hire senior security engineers to make. The framework converts that judgement into a values problem. That's the move to watch for, and it isn't unique to AI: **a technical disagreement reclassified as a loyalty question stops being answerable with evidence.** Once "I don't think this is reliable enough for security work" is heard as "you're fighting the company," there is no finding that could settle it. The engineer who was right for good reasons and the engineer who was merely obstinate are now in the same bucket. Worth noting the flip side honestly: an organization genuinely does need to be able to change direction, and a person who will not engage with a strategic shift at all is a real problem. The speaker isn't describing something unheard of. He's describing it *out loud*, which most wouldn't. Read alongside the sibling clip where the same leadership says *the CEO goes first* and made sure the first hour with the tools was good. That's the carrot. This is the stick, and the set is more honest for containing both. # ✅ Action Items **DevRel** - If advocating AI tooling internally, keep the evidence channel open — mandates that classify dissent as disloyalty lose the feedback that would tell you the tool is failing. **OSPO** - Any AI-adoption policy for contributors should distinguish "hasn't tried" from "tried and judged it unsuitable," and preserve a route for the second. **Community** - Volunteer communities cannot use this enforcement model at all; the analogous failure is quieter attrition. Worth naming explicitly in contributor guidance. **WordPress subteams** - **Core AI:** as AI tooling enters core workflows, ensure technical objections stay technically answerable rather than becoming position-taking. - **Core / Core Program Team:** decision records should capture dissenting technical reasoning, not just outcomes. - **Community / Training:** relevant to how contributor onboarding frames tool adoption. - **Support / Documentation / Design / Marketing / Meta / Mobile / Playground / Photos / TV / Polyglots / Accessibility / CLI / Hosting / Tide / Openverse / Themes / Plugins / Core Performance:** no direct action from this source. - **The WPOCC:** directly relevant to employment guidance for member agencies adopting AI tooling. - **FAIR:** no direct action from this source. # 🧪 Hypotheses 1. Organizations that classify AI non-adoption as a values issue show higher voluntary turnover among senior technical staff than those treating it as a training issue. 2. Reclassifying technical dissent as loyalty reduces the rate at which tool deficiencies are reported upward. 3. Employees who "refuse to use AI for security work" hold measurably different reliability assessments than those who adopt, testable against actual tool error rates. 4. Adoption mandates paired with visible leadership use (the sibling clip's approach) achieve higher voluntary adoption than mandates alone. 5. Departures framed as employee-initiated in mandate contexts correlate with prior performance-management contact. # 📊 Methodology **None** — a ~60-second excerpt describing what an organization did, with no data attached. Nothing quantified: not how many people were at level 0, how many left, what the maturity scale actually measures, how placement on it was determined, or who did the determining. The scale itself is referenced as though the audience knows it, because in the parent talk they do. Critically, the account is **entirely from the party that made the decisions**. No departed employee's reasoning is represented, and their objections are characterised only in the leadership's framing ("dismiss AI as hype"). Whether any of them had a defensible technical case is unknowable from this source — which is exactly why the reclassification is worth flagging rather than accepting. The speaker's candour is real and worth crediting: this is a more honest account than most organizations give of the same events. Candour about a decision is not evidence that the decision was correct. # ❓ Questions - What actually distinguishes level 0 from level 1 operationally — zero usage, or expressed disagreement? - Were any level-0 objections evaluated on their technical merits before being classified as principled opposition? - How many people were affected, and at what seniority? - Did anyone move from level 0 to engaged, and what changed for them? - Is "refuses to use AI for security work" ever a correct professional judgement in this framework, and if so how would it be recognised? - What feedback channel survived, for reporting that a mandated tool is failing? # 🔮 Future Research - Outcome study of AI-adoption mandates versus voluntary programmes on retention and on realized productivity. - Research on how reclassifying technical dissent affects upward reporting of tool failures. - Longitudinal work on whether early AI refusers in security contexts were vindicated or not. - Comparative study of carrot-only, stick-only, and combined adoption programmes. # 💡 Implications **Practice.** The reclassification is the mechanism worth naming: Levels 1–3 are treated as a skill gap needing time, while Level 0 — "not engaged" — is reclassified as a *values* disagreement. The move converts a training problem into a conduct problem, and the two have very different consequences. **Policy.** Escalating non-adoption to HR, with some of those people subsequently leaving, is an employment consequence attached to tool use. Whether that is defensible depends on whether the tool is genuinely load-bearing for the role — but the decision is being made through a values frame that does not require answering that question. **Security.** Not applicable to this source. **Education.** This clip retroactively changes how the rest of the set reads. Advice about giving agents a performance review lands differently once it is clear the humans were also being assessed on adoption. **Funding.** Not applicable to this source. # 🧠 Learning Styles Assessment - **Visual:** Minimal — talking-head clip. The maturity scale is referenced verbally and would benefit enormously from the slide it presumably had. - **Aural:** Adequate; conversational, with the heaviest disfluency of the six clips concentrated on the departures sentence. - **Verbal:** Moderate. "Level zero," "not engaged," and "skill issue" are used as defined terms without definition in this excerpt. - **Kinesthetic:** Not applicable. - **Logical:** Structurally clean — a binary classification with different consequences per branch. Sound as a structure; the question is whether the classification is valid. - **Social:** Low as delivered, though the content is entirely about organizational dynamics. - **Solitary:** Limited — no framework or reference to follow up. - **Naturalistic:** Not applicable to this source. **Reading level estimate:** roughly grade 10–12 equivalent. Plain language throughout; the difficulty is not vocabulary but the missing context of the maturity scale, which the clip assumes. # 📊 Accessibility Assessment Mapped to WCAG 2.2 AA: 1. **Captions (1.2.2):** Auto-captions present and accurate — plain vocabulary with no proper nouns to mangle. Pass. 2. **Audio description (1.2.5):** Not applicable in practice; content is spoken. If the parent talk showed the maturity scale as a slide, that context is lost in the clip for all viewers equally. 3. **Keyboard/navigation (2.1):** Platform-determined. Pass. 4. **Timing / seizure safety (2.3):** No flashing content. Pass. 5. **Reading level (3.1.5):** Passes on language; fails on assumed context. 6. **Predictable structure (3.2):** Good — two branches, stated in order, with consequences attached. 7. **Assistive-tech compatibility (4.1):** Standard platform; transcript machine-readable. Pass. **Note:** the clip is decontextualised by editing. A listener cannot tell what levels 0–3 measure, which materially affects whether the classification sounds reasonable. # 🌍 DEIB Assessment - **Cultural breadth:** Not directly addressed. Worth noting that mandate-driven adoption programmes interact with visa-dependent and otherwise less-mobile employees very differently than with those who can leave freely — "there were people that left the company" is a lower-cost outcome for some staff than others, and the clip does not consider this. - **Inclusive language:** Mostly neutral. "Skill issue" carries dismissive connotations from internet usage, though it is used here in a genuinely charitable sense — the people it describes get time and support. The framing "fighting back against the company" is the loaded phrase. - **Neurodiversity support:** Not addressed, and relevant. A maturity scale measuring engagement with a new tool over a fixed period will register differently for people who adopt tools at different rates or who need more structured onboarding. Levels 1–3 accommodate this well ("more time to play"); the level-0 classification does not obviously distinguish disengagement from difficulty getting started. - **Representation:** Single speaker, leadership perspective only. The departed employees are described but not heard, and their reasoning appears only as characterised by the person who classified it as opposition. This is the most significant representational gap in the clip. - **Psychological safety:** **The weakest dimension, and the substance of the clip.** A policy that treats a stated technical position as evidence of misalignment with company direction reduces the safety of voicing technical concerns generally. The specific example — refusing AI for security work — is a judgement call in the company's core domain, and classifying it as a values problem tells everyone watching what the cost of that judgement is. Against this: levels 1–3 are handled with genuine patience, and the sibling clip's leadership-goes-first approach suggests the programme was not purely coercive. # ✅ Logical Validity ## Strong reasoning - **The skill/will distinction is a real and useful one.** Not being able to and not agreeing to are genuinely different problems requiring different responses, and conflating them is a common management error. The clip avoids it. - **Consistency of treatment within the skill branch.** Levels 1–3 all get the same response — time — with no escalating pressure by level. That is coherent. - **The candour is itself a form of rigour.** Stating plainly that HR was involved and that people left, rather than describing "a natural evolution of the team," makes the claim checkable and the reasoning inspectable. - **The organizational premise is defensible.** A company that has committed to a direction does need people to engage with it, and sustained total non-engagement with a strategic shift is a legitimate management concern. ## Fallacies 1. **False dichotomy — the central flaw.** The framework offers exactly two explanations for non-adoption: insufficient practice, or opposition to the company. It has no category for **tried it, evaluated it, and concluded it is unsuitable for this work** — which, in a security firm, is a conclusion a competent engineer might reach on evidence. **Fix:** add an "informed dissent" branch with an evidentiary process. 2. **Question-begging on the examples.** "If you dismiss AI as hype" assumes the dismissal is unreasoned. Someone who has evaluated the tooling against their domain and found it wanting is not dismissing hype; they are reporting a result. The classification treats the conclusion as proof of the attitude. 3. **Loyalty framing forecloses evidence.** Once a position is "a disagreement on principles," no technical finding can resolve it — principles are not settled by benchmarks. This is the mechanism by which the reclassification removes the objection from the domain where it could be answered. Note the sibling clip's insistence on benchmarking agents against existing tools: **the company built an evidentiary process for evaluating the agents, and a loyalty process for evaluating the humans who doubted them.** 4. **Agency displacement.** "There were people that left the company because they were stuck at that stage" attributes the departure to the employee's position rather than to the decisions made about them following HR involvement. The passive construction does real work. 5. **Survivorship in the evidence base.** The people who remained are the sample from which the programme's success will be judged. Those who objected are gone, and with them whatever their objections were tracking. 6. **Unfalsifiable classification.** No criterion is offered by which someone could demonstrate that their non-adoption was reasoned rather than resistant. Without one, the category cannot be exited by argument — only by compliance. ## Tone of voice Matter-of-fact and notably unapologetic, delivered at the same brisk pace as the technical clips in the set. There is no euphemism — HR is named, departures are named — and no defensiveness either. The disfluency spike on the departures sentence is the one place the delivery breaks from the rest. It would be over-reading to call it discomfort from a transcript alone, but it is the sentence where the fluency goes. What is absent is any register of cost. The people who left are described without evaluation of whether the company lost anything, whether any of them were right, or whether the outcome was regrettable. The clip treats it as a resolved administrative matter, and that flatness is itself the most revealing thing about the register. # 🛡️ DARVO Analysis Not a defensive response to accusation, so no DARVO in the classic sense. But this is the one clip in the set where the components are worth examining seriously rather than dismissing. **Deny.** Not present. The speaker does not minimise what happened — he states it directly, which is the opposite of denial. **Attack.** Present in a structural rather than personal form. The departed employees are characterised as "fighting back against the company" and as not having "adopted the vision." No individual is named or derided, but the group is assigned an adversarial posture that they may not have claimed for themselves. The characterisation is unilateral: they are not present to describe their own position. **Reverse Victim & Offender.** **This is the component that genuinely applies.** The organization holds all the power in the relationship — it sets the mandate, defines the scale, assigns the level, and involves HR. Yet the framing positions the *company* as the party being resisted: employees are "fighting back against the company," which casts the institution as the object of aggression and the individual as its source. The party with the power to end the employment relationship is rhetorically the one under attack. That inversion is not necessarily intentional, and it is common in organizational language. But it is the same structure identified in the Erika Kirk and Hebrew Roots material processed in this batch: **an entity with structural power adopting the position of the party being acted upon**, which converts objection into aggression and makes the objector the offender. **Verdict:** No deny. Structural attack via unilateral characterisation. One genuine victim-offender inversion in the "fighting back against the company" framing — worth recording because it is a *secular, corporate* instance of a pattern this batch has now documented in two religious contexts, which strengthens the general claim considerably. # 🎯 Conclusion **Bottom line.** The clip's real content is a classification, and the classification is where the problem lives. Non-adoption is sorted into *needs practice* (patient, supported, no consequence) or *opposes the company* (HR, departures). There is no third category for **evaluated it and judged it unsuitable** — and the two examples given of level-0 behaviour are both positions rather than refusals, one of them a professional judgement about tool suitability in the firm's own domain. The mechanism to carry forward: **reclassifying a technical disagreement as a loyalty question removes it from the domain where evidence could settle it.** The sibling clip's insistence on rigorous agent benchmarking makes the contrast sharp — an evidentiary process for the tools, a values process for the people who doubted them. **What to hold loosely.** Nothing is quantified, the maturity scale is never defined in the clip, and the entire account comes from the party that made the decisions. Whether any departed employee had a defensible case is unknowable here. The organizational premise is also genuinely defensible: companies do need to be able to change direction, and total non-engagement is a real problem. **Next steps.** Find the parent talk for the maturity scale definition. Read as a set with the other five clips — the leadership-goes-first material is the same programme's other half, and the set is more honest than either half alone. If you are designing an adoption programme: write the informed-dissent branch before you need it, and decide what evidence would let someone exit the category by argument rather than by compliance. # 🧱 Refactor Appendix ## Literature Split Pad ### Non-engagement is reclassified from skill gap to values opposition > [!quote] > If you&#39;re in level zero though, if you&#39;re not engaged, that means you&#39;re fighting back against the company. It means that uh, you are you have not adopted the vision and the direction that the company is now going in. *Where:* ~00:00:00. *Why:* the pivot the whole policy turns on. Everything below this line gets patience; everything at it gets HR. ### The examples of "not engaged" are positions, not refusals to try > [!quote] > if you dismiss AI as hype, if you refuse to use AI for security work. This is a disagreement on principles, not on skills. *Where:* ~00:00:00. *Why:* the second example is a professional judgement about tool suitability in the company's core domain — exactly what senior security staff are hired to make. The framework has no category for it being correct. ### The consequence is stated plainly > [!quote] > for people that were stuck in the not engaged category, we had hard conversations with our HR team, and uh, there were people that left the company *Where:* ~00:00:31. *Why:* candour worth crediting — most organizations describe this as attrition. Note the passive framing that locates cause in the employee's "stage." ### Skill gaps get time > [!quote] > Levels 1, 2, and 3 are a skill issue. Just means you haven&#39;t spent enough time with the tool to get good at it. *Where:* ~00:00:00. *Why:* the contrast that defines the policy — patience for incapacity, escalation for disagreement. ## Atomic Split Pad #### Reclassifying a technical disagreement as a loyalty question removes it from the reach of evidence Principles are not settled by benchmarks, so recasting an empirical objection as a values position makes it unanswerable rather than answered. <!-- atomicity: 1 idea ✓ --> #### A framework that cannot represent informed dissent converts correct objections into misconduct When the only categories are insufficient practice and opposition, the person who evaluated the tool and was right is indistinguishable from the person who refused to look. <!-- atomicity: 1 idea ✓ --> #### An organization can hold all the power and still occupy the position of the party resisted "Fighting back against the company" casts the institution as the object of aggression, inverting who acts on whom. <!-- atomicity: 1 idea ✓ --> #### Selecting out the doubters removes the population whose objections would have measured the decision The remaining sample cannot falsify the strategy, because the people tracking its failure modes are gone. <!-- atomicity: 1 idea ✓ --> ## Fleeting Queue - #xref **Third instance in this batch of a powerful party occupying the victim position** — with the Erika Kirk deep-read ("enemy occupied territory" from a nationally resourced organization) and the Hebrew Roots episode. This one is *secular and corporate*, which substantially strengthens the general claim. Strong Permanent candidate; the cross-domain spread is the evidence. - #xref Sharp internal contrast with the sibling clip [[Give Your Agents a Performance Review]] — rigorous evidentiary benchmarking for the agents, loyalty classification for the humans who doubted them. Worth one note holding both. - #xref "Distress/dissent reframed as opposition" is structurally the same as the distress-reframing atom from the Hebrew Roots and Erika Kirk sources — internal signal relocated into external conflict. - #xref The unfalsifiable-category point connects to [[Validation That Inspects Steps Cannot See Composition]] only loosely, but strongly to any note on self-sealing frameworks. - #articleseed "Write the informed-dissent branch before you need it" — for anyone designing a tool-adoption programme. - Vocabulary: **adoption maturity scale**, **not engaged**, **skill issue**, **principled disagreement**. - Lingering question: was any level-0 objection ever evaluated on its technical merits? The clip gives no criterion for exiting the category by argument. - ⚠️ Sourcing caution: entirely the decision-makers' account. Departed employees are characterised, never heard. ## Things 3 - Find the parent talk to get the maturity scale definition - Check whether Trail of Bits has written publicly about the adoption programme <!-- deep-read:sections end --> 📄 **Full text:** [[Full Document Contents/Articles/A Disagreement on Principles, Not on Skills.md|A Disagreement on Principles, Not on Skills by YouTube]] ## Highlights - Levels 1, 2, and 3 are a skill issue. Just means you haven't spent enough time with the tool to get good at it. You know, if you're capable, means that you just had a few hours of experience. Um, and you need more time to experiment to get to level two. If you're in level zero though, if you're not engaged, that means you're fighting back against the company. It means that uh, you are you have not adopted the vision and the direction that the company is now going in. Uh, if you dismiss AI as hype, if you refuse to use AI for security work. This ([View Highlight](https://read.readwise.io/read/01kywq0z4vt3ns03c9vgsh384f)) - Note: The pivot the policy turns on. Note both level-0 examples are POSITIONS, not refusals to try — and "refuses to use AI for security work" is a professional judgement about tool suitability in the firm's own domain. The framework has no category for that judgement being correct, so it can't be exited by argument, only by compliance. - Tags: technology-mandates, atom/literature/h1, dissent - is a disagreement on principles, not on skills. So, for people that were stuck in the not engaged category, we had hard conversations with our HR team, and uh, there were people that left the company uh, because they were um, stuck at that stage. Uh, but for anybody that's at level 1, 2, or 3, we just treat this as like, "Okay, you know, you need more time to play." ([View Highlight](https://read.readwise.io/read/01kywq1bg8316sxqpy8rk01csd)) - Note: Candour worth crediting — most organizations call this attrition. Note the passive framing locating cause in the employee's "stage" rather than in the decisions made about them. Also the asymmetry that IS the policy: incapacity gets time, disagreement gets HR. - Tags: atom/literature/h1, organizational-change, workforce-management