AI Enthusiasts Are in a Race Against Time, AI Skeptics Are in a Race Against Entropy

Charity Majors · rss · 2026-07-11 Reader 🖥️

practical · Persuasive practical essay (leadership argument) — a book of rules for how to act, not a theory of how things are. — A prescriptive, diagnosis-then-prescription essay by Honeycomb's CTO aimed at engineering leaders and staff+ engineers navigating the AI transition. It argues from a single composite cautionary tale plus one exemplar (Fin/Intercom) rather than systematic evidence, and is deliberately warm, profane and metaphor-heavy. Its aim is behavioral change — to reconnect two feuding camps — not to establish a proposition about the world.

Structure

Key terms

• race against time
  The enthusiast's existential threat: teams that sit out AI while competitors lean in hard could be out of business before the dust settles — the window, not the technology, is the risk.
• race against entropy
  The skeptic's existential threat: shipping code faster than anyone can read it degrades reliability, evaporates institutional knowledge and leaves systems nobody understands. Majors never uses the word 'entropy' in the body — she describes it ('reliability degrades... products burbling into incoherence... systems melting into slop'); the title carries the term.
• no natural feedback loop
  The structural diagnosis: the wins and the costs of AI acceleration accrue to two different groups, so the person who reaps the win never feels the downstream cost and cause is disconnected from effect.
• tell the whole story
  The first remedy: whoever celebrates a win must also surface its downstream costs (and whoever surfaces costs must include what was achieved), in the same forum where the win was announced.
• AI is an amplifier
  Borrowed from Nathen Harvey / the 2025 DORA report: AI magnifies the strengths of high-performing organizations and the dysfunctions of struggling ones — so it cannot be a shortcut around engineering discipline.
• moral authority of ownership
  The standing to decide and to push back is earned by owning the consequences: whoever is left holding the bag should get final say over what goes in it, so engineers who run the code must be extremely involved in defining what they agree to support.
• credible authority (on AI)
  For a critique to land, the critic must be a genuine expert in the thing being critiqued; arguing against the new ways purely from a position steeped in the old ways forfeits standing.
• Scylla and Charybdis / don't sink the boat
  The two symmetric failure modes leaders must steer between — the Scylla of missed windows and the Charybdis of systems melting into slop; the leader's job #1 is not to sink the boat while threading between them.

Propositions

• explicitdevelopedThe AI enthusiast/skeptic divide in high-performing engineering orgs is not a case of one side being right and the other deluded; each camp is grappling with a real, escalating, existential threat to the company, and the closer each looks the more real evidence it finds.conf: high
  extends ←Stop being skeptical about AI for development with Charity Majors · p13 — Reprises, nine months later and in a different venue (interview vs. essay), Majors's own earlier two-camps-both-right thesis from her essay already in the corpus.
• explicitThe enthusiasts are not wrong: teams that lean hard into AI are producing real, non-imaginary, discontinuous leaps in capability, and this is not a normal cycle you can wait out — teams that sit it out could be out of business before the dust settles.conf: high
• explicitdevelopedThe skeptics are not wrong: shipping code faster than engineers can read it, in domains nobody fully understands, makes withdrawals from a trust account that took years to build — reliability degrades, institutional knowledge evaporates, and you get systems nobody understands and on-call rotations that grind people up.conf: high
  supportsPlease stop the AI Confidence Theater · p7 — Identical trust-account-withdrawal metaphor: each inflated AI claim is a withdrawal from a slow-built, shared trust balance.
  extends ←All models, no matter how smart, will eventually build systems... · p1 — Majors's p3 locates the trust/comprehension breakdown in human engineers unable to read AI-shipped code fast enough; Yegge's tweet extends the same entropy-outrunning-comprehension pattern to the model's own comprehension of its own output, not just the human reviewer's.
• explicitThe wins and the costs of AI acceleration happen to two different groups of people, so there is no natural feedback loop connecting the person who reaps a win with the person who pays its downstream cost.conf: high
  refinesPodcast: The AI Joy Gap: Why Some Developers Thrive While Others Struggle · p1 — Relocates the polarization from a task-type story (greenfield gains vs. legacy friction) to a feedback-asymmetry story: the divide is driven by wins and costs landing on different people, not just by what kind of code each camp touches.
• explicitBecause downstream costs are usually invisible to the person who incurs the win, enthusiasts sincerely under-report costs (all they see is the win) while skeptics, hearing ever-bigger and less-tethered claims, respond with escalating cynicism until they no longer believe a word enthusiasts say.conf: high
  supportsAI productivity is burning out your best engineers · p1 — Majors's 'downstream costs are invisible to the person who reaps the win' is the general form of the 'invisible validator' work that subsidizes others' AI productivity while never appearing on any dashboard.
• explicitThe information asymmetry is structural rather than malicious: wins get trumpeted in public settings (blog posts, conference talks, all-hands) while costs bubble up in private ones (SRE meetings, on-call, retros, DMs), so both camps can simultaneously and sincerely feel they are the ones being unfairly silenced.conf: high
  supportsPodcast: The AI Joy Gap: Why Some Developers Thrive While Others Struggle · p7 — Her structural forum-asymmetry claim explains the observed disconnect between leadership's AI hype and developers' on-the-ground reality that breeds distrust from both sides.
• explicitReal, discontinuous AI capability gains are achievable for teams founded pre-AI, but they happen because of pre-existing engineering discipline — high discipline, fast feedback loops, and a culture of experimentation and measurement — not because AI is magic (the Fin/Intercom case: 3x output in 9 months, defect backlog halved, downtime down 35%).conf: high
• explicitdevelopedThe first fix is to mend shared reality by telling the whole story — pairing every celebrated win with an honest account of its downstream costs, inviting feedback across power dynamics, and closing the loop back to whoever caused a cost — which acts as a steam valve for anger, makes people feel seen, and bends toward cheaper wins.conf: high
  extends ←Stop being skeptical about AI for development with Charity Majors · p13 — Same 'tell the whole story, pair wins with costs' prescription, restated as a direct appeal to both camps in interview form.
• explicitThe second fix is to treat AI-adoption disputes as engineering problems, not epistemological debates: replace 'what's wrong with you' and 'that will never work' with 'what would it take for you to feel comfortable shipping code to production without reading it?' and put the answer — better evals, tests, feature flags, guardrails, observability, decoupling, blast-radius reduction — on the roadmap.conf: high
  supportsThe Organization Is the Bottleneck · p10 — Treating adoption as an engineering problem — with guardrails and blast-radius reduction on the roadmap — is the operational form of 'if we ship 3x faster, deployment safety nets become even more important.'
• explicitestablishedAI is an amplifier that magnifies the strengths of high-performing organizations and the dysfunctions of struggling ones, so it is not a replacement for or a shortcut to engineering discipline — which makes discipline more vital, not less.conf: high
  supports11) AI-spesial med Vilde Opsal og Anders Sveen · p5 — Both invoke the same DORA/Nathen Harvey finding that AI is a force multiplier in both directions — strong teams get better, weak teams get worse faster.
  supportsAI Is Changing Software Engineering. The Fundamentals Still Matter. · p15 — Same conclusion in different words: amplification turns previously-optional engineering discipline into load-bearing infrastructure.
  supports ←Stop being skeptical about AI for development with Charity Majors · p10 — Same practitioner's claim that AI raises rather than lowers the required discipline bar, restated here specifically for non-deterministic/AI-written systems.
  supports ←AI-Assisted Software Development: Team Profiles and Capabilities for Putting Research into Action · p1 — The identical Harvey/2025-DORA amplifier quote is reused as evidence in a different secondary piece — same primary claim, independent framing.
• explicitMoral authority and the standing to push back are earned by owning the consequences: whoever is left holding the bag should get final say over what goes in it, so engineers who own and operate code must be at minimum extremely involved in defining the conditions under which they agree to support it.conf: high
  supportsResearch: Why You Shouldn't Treat AI Agents Like Employees · p6 — Her 'whoever holds the bag gets final say' rests on the same premise — AI cannot be held accountable, so clear human ownership is required.
  supportsPodcast: The AI Joy Gap: Why Some Developers Thrive While Others Struggle · p12 — Reinforces that developers remain personally accountable for code they submit regardless of whether AI generated it.
• explicitA critic's pushback only lands if the critic is a credible authority on the topic — AI itself; arguing against the new ways from a position steeped only in the old ways forfeits standing, so skeptics should be highly motivated to ground themselves in genuine expert knowledge of AI.conf: high
  supportsFragments: February 13 · p3 — Her prescription that skeptics build real AI expertise to earn standing rests on the empirical mechanism that hands-on practical experience is what converts skeptics by giving them credible grounds to judge value.
• implicitRebuilding a shared reality between the camps is the precondition for any technical solution: the productive engineering conversation is impossible until both sides believe they inhabit the same reality, share some goals, and treat each other as reasonable actors capable of changing their minds — you need not agree on who is right.conf: high
• explicitThe defining leadership challenge of the moment is steering between two symmetric failure modes — being late to AI and grinding the team into a pulp — and leaders (including title-less staff+ engineers) must bring people along by understanding their pushback rather than forcing decisions; forcing something through is a last resort, and when a leader does lay down the law they had better be right.conf: high

Arguments

Narratives

• A conference presenter tells an easy-breezy story of astonishing vibe-coding wins (year-long rewrites done in weeks, backlogs cleared) to a room of enthusiasts who eat it up uncritically — while colleagues at the same company describe those projects as a horror show, with teams still grinding through cleanup months later. Majors flags it as a deliberate composite ('I am subtweeting MANY TALKS').
  → Opening exhibit: dramatizes the selective success story (confidence theater) and the invisible-downstream-cost asymmetry that the whole essay diagnoses.
• The Fin (formerly Intercom) engineering org: CTO Darragh Curran challenged R&D to double productivity in a year ('2x'); nine months later they reported 3x'ing output (merged PRs / R&D headcount) with mixed but real results — defect backlog more than halved, >2x product changes, 39% faster idea-to-shipped, code quality improving after 18 months of decline, downtime down 35%. Footnotes concede special conditions: a decade of in-house AI expertise and a near-death experience that 'cleared the deck.'
  → Majors's 'north star of good' — her existence proof answering 'is any of it real?', while modeling the honesty (PRs are imperfect; special conditions) she prescribes.
• A caricatured exchange: an enthusiast demands shipping without code review ('Company X is doing it... Why do you hate the future?') and the skeptic explodes ('you want me to just auto-accept this shit?... your mother had a face like a donkey'). Both are partly right; the productive version replaces the insults with 'What would it take for you to feel comfortable shipping code to production without reading it?'
  → Dramatizes the difference between rhetorical conversation-stoppers and the engineering reframe that is the essay's second prescription.
• An exasperated engineer tells Majors: 'I don't WANT to be an AI hater. I studied AI in school! ... I feel like I'm getting backed into a corner where I have to be a hater because I'm the only one left who gives a shit about reality! Is any of it real?'
  → Humanizes the skeptic's trap and motivates Majors's decision to 'show her work' with the Fin example.

Questions addressed

• ✅ Is any of the AI productivity story real, or is it all hype?
  Partly real: the Fin case is a genuine discontinuous leap, but it was gated on pre-existing engineering discipline and a measurement culture — it is not magic, and PR-count is an admittedly imperfect proxy.
• ✅ Why have enthusiasts and skeptics stopped talking, and why does the divide keep widening?
  Because the wins and costs accrue to different people so there is no natural feedback loop, and because the asymmetry of where wins vs. costs surface is structural rather than malicious.
• ✅ How should teams settle concrete disputes like 'ship without code review'?
  Reframe them as engineering problems — ask 'what would it take?' and put the prerequisites (evals, tests, guardrails, observability, blast-radius reduction) on the roadmap — rather than as epistemological debates.
• ✅ Who has the standing to push back on how AI gets used?
  Those who own and operate the consequences AND who are credible authorities on AI itself; ownership grants the right to be heard, expertise makes the critique land.
• ✅ Is AI a shortcut around engineering discipline?
  No — AI is an amplifier of existing strengths and dysfunctions, so engineering discipline becomes more vital, not less.

Gaps

• incompleteThe essay explicitly scopes itself to relatively high-performing teams that already have engineering discipline, fast feedback loops and a measurement culture, and excludes disconnected management, low-discipline/low-trust teams, tiny startups and behemoths. But that is most organizations. The whole 'apply engineering rigor / what would it take?' prescription presupposes the very prerequisites the excluded majority lack, and Majors offers no path for how a team without Fin-level foundations bootstraps them under competitive pressure.
  Read next: Read enterprise-AI readiness ladders and the 'velocity multiplier becomes a debt accelerator' work for teams without foundations: how low-discipline orgs sequence guardrails, CI/CD, testing and observability before leaning into AI.
• incompleteThe core remedy — 'tell the whole story,' 'invite feedback,' 'close the loop,' 'make it easy' — is a behavioral/individual-virtue prescription, yet Majors herself names the structural incentives working against it (wins are trumpeted publicly, costs griped privately; real power dynamics make honest cost-disclosure risky). Beyond 'surface it in the same forum,' she gives no incentive or mechanism-design fix that would make disclosing your own downstream costs structurally rewarded rather than merely encouraged.
  Read next: Read commons-governance and mechanism/costly-signaling work (Ostrom on shared resources; signaling under cheap talk) and psychological-safety instrumentation: how to make cost-disclosure a low-cost, high-status default rather than an act of individual courage.
• incompleteThe 'north star of good' rests on a single, self-selected success case (Fin) whose own footnotes concede special conditions — a decade of in-house AI expertise and a near-death experience that 'cleared the deck' — and whose headline metric (merged PRs / R&D headcount) she admits is imperfect. How representative or reproducible the 3x result is for teams lacking those conditions is asserted ('well within reach for the rest of us') more than demonstrated.
  Read next: Read controlled/longitudinal AI-productivity studies and the measuring-AI-impact literature on selection bias in success case studies and on outcome-anchored, oppositional metrics beyond PR counts.
• unstated-assumptionThe framing quietly grants the enthusiast premise as more-likely-true — that the field is 'directionally moving toward software factories and AI-validated diffs' and that the capability leap is durable — which makes the 'race against time' threat as real as the 'race against entropy' one. The essay never entertains that the enthusiast direction could itself be partly hype or bubble, which would deflate the missed-window threat and unbalance the two-sided symmetry it depends on.
  Read next: Read the AI-bubble and frontier-lab-hype debates to test whether the 'missed window' existential threat is as well-evidenced as the reliability-decay one, or whether the symmetry is partly assumed.

Related documents

• The Organization Is the Bottleneck
• Podcast: The AI Joy Gap: Why Some Developers Thrive While Others Struggle
• Research: Why You Shouldn't Treat AI Agents Like Employees
• AI productivity is burning out your best engineers
• AI Is Changing Software Engineering. The Fundamentals Still Matter.
• 11) AI-spesial med Vilde Opsal og Anders Sveen
• The 8 software engineering metrics AI broke
  Cognitive Debt Measuring AI Impact Technical Debt
• AI coding creates two kinds of debt. You're only measuring one
  Cognitive Debt Measuring AI Impact Technical Debt
• Presentation: AI-First Software Delivery: Balancing Innovation with Proven Practices
  AI as Amplifier Feedback Loops Vibe Coding
• Software Development As Collective Learning With Hazel Weakly
  AI as Amplifier Feedback Loops Psychological Safety