Itchy Brain (Still Burning: Michael Grinich & Kent Beck on AI adoption, enterprise developer tools)

Kent Beck — Still Burning Podcast · podcast · 2026-05-20 Reader 🖥️

AI Labor Disruption Emotional Impact of AI Work Enterprise AI Adoption & Readiness Explore / Expand / Extract Hands-On Leadership & the Pendulum Agency Industrialization of AI Intrinsic Motivation & Making Measuring AI Impact Near-Zero-Cost Code

We are not in another technology hype wave but in a step-function transformation: the marginal cost of producing software is collapsing toward zero, large incumbents are adopting AI faster than any prior shift, and the whole ecosystem — not just the AI labs — is accelerating; in this 'Red Queen' regime, profit and productivity may not visibly improve because every competitor is running at the new tempo, so the only sane response is to paddle straight into the wave by sustaining the maker's compulsion (the 'itchy brain'), reshaping engineering leadership toward hands-on impact, and accepting that software engineering as a profession will resemble medicine or law more than the unique 2010s template that minted six-figure undergrads.

mixed · mixed — A long-form podcast conversation between an experienced craft thinker (Beck) and an enterprise-infrastructure founder (Grinich). Practical in its diagnoses (what AI is doing to engineering management, what 'enterprise ready' means, how big companies are adopting) and theoretical in its framing (the Red Queen evolutionary analogy, the marginal cost of software approaching zero, software engineering as a historically anomalous profession). Form is mixed: expository (WorkOS's vantage point, Grinich's biography), argumentative (why this AI moment is a step function, not just acceleration), and narrative (the Himalayan climber-Sherpa story, Beck's Gusto payroll example, Grinich's 18-month founder gap-year, Beck's reflection on JUnit and scaling). The conversation does not have a single explicit thesis; the joint thesis is constructed across both speakers.

Structure

Part	Role
Opening sponsor frame: agents do the watchmaker work; trust/identity is the remaining gap	Beck-as-host sets the conceptual stage: the gap from idea to running system has narrowed, but the gap from running system to enterprise-trusted system has not. Names the WorkOS problem.
Grinich's introduction and the 'picks and shovels' vantage point	Establishes Grinich's epistemic position: WorkOS sees the go-to-market and infrastructure adoption patterns across hundreds of SaaS and AI companies. He sells to both 'previous era' companies (Vercel, Webflow, Carta) and 'AI era' companies (ChatGPT, Claude, Cursor, Perplexity).
The three-fold acceleration thesis: faster build, better products, big-company appetite	Grinich's central diagnostic argument — code agents compress timelines, AI makes products more human, and large enterprises (catalyzed by their COVID transformation muscle memory) are now adopting fast. Capital amplifies all three.
Beck's biographical detour: Gusto's payroll signal as an analogy	Beck checks the claim against his own prior vantage: at Gusto, payroll movement leaked macro shifts. He asks Grinich whether the second derivative — growth of growth — is positive. It is.
Grinich's biography: McMinnville, the realization that the world is malleable, software as highest-leverage medium	Personal narrative supporting the 'agency' theme: ideas live or die based on whether you believe the world is something you can change. Software wins on zero marginal distribution cost.
The itchy brain: maker compulsion as the only durable founder motivation	Beck names the felt phenomenon ('itchy brain' — the idea won't leave until you make the thing). Grinich extends it: external-validation motivation burns out; the durable kind is wanting to make stuff and connect with people. Illustrated with the Himalayan climber-Sherpa parable (look at the step in front of you, not the summit).
Three phases of a company: explore, expand, extract	Beck's framework. Each phase has incompatible skill demands; cognitive biases retcon the expand phase into either inevitability or near-disaster. Grinich confirms it feels like a different company every six months.
Engineering management is changing: the hands-off people manager is evaporating	Grinich's claim: pure people managers who delegate and create psychological safety are losing ground to leaders driving customer value and product shipping. CTOs are returning to code. The same pattern may spread to finance, legal, sales, support.
Beck's worry: where does coaching/mentorship go?	Beck pushes back from craft territory: the weekly 1:1, the pattern-spotting mentor was the apprenticeship layer. If managers now have 50 reports, that infrastructure disappears.
Software engineering as historically anomalous: the 'four-year undergrad to six figures' template	Grinich's reframe: no other professional discipline (medicine, law, architecture, civil/chemical/mechanical engineering) lets a fresh graduate command Meta L3 wages. Software was the anomaly; it will normalize toward residencies, fellowships, specialization, smaller in-house teams of higher-paid specialists.
Beck's counter-trend: the descaling of software	Beck names a colliding trend: the barrier to getting something working is dropping, so less-skilled people will increasingly make decisions formerly reserved for the more skilled. Beck does not resolve the tug of war.
The marginal cost of software → zero and the measurement problem	Beck argues that 'shipped = done' accounting is incomplete: operation cost, error rate, business cost of errors all remain. He points to profit as the honest measure of AI impact. Grinich counters: not profit but speed — because all competitors are using AI, profit baselining is impossible.
The Red Queen framing	Grinich introduces the Red Queen evolutionary theory: in nature every animal is at its survival limit, no slack, all running at the same pace just to exist. By analogy, AI adoption will not show up as profit growth — it will show up as faster experimentation, faster market exploration, increased competitive intensity. Falling off the pace means falling away.
Industrialization-phase analogy and the closing 'paddle into the wave'	Grinich's optimist framing: we are in the dirty-transformation phase (token budgets blown by May, no baseline yet, just spend what it takes). The goal is to get to the other side of the change as fast as possible. The kayaking metaphor: head straight into the wave; sideways gets you flipped.
Grinich's worry: white-collar labor disruption is real, fatigue makes people miss it	His one named fear: people have crypto-fatigue and VR-fatigue and may misread this as another hype wave, when at its core it is a labor disruption (he names Amodei). Not catastrophic, but a big shock.

Key terms

Red Queen (evolutionary theory) Industrialization of AI
Grinich's framing device, drawn from evolutionary biology via Alice in Wonderland: every species is already at its survival limit; everyone runs as fast as they can just to stay in place. Applied to AI adoption, it means the whole ecosystem will accelerate together, so profit and productivity gains may be invisible at the company level — falling behind means falling away.
marginal cost of (creating) software → zero Near-Zero-Cost Code
Grinich's claim that the cost of writing code has 'gone essentially to zero' with coding agents. Specifically the marginal cost of producing one more unit of software output, not the total cost of operating software systems. Beck immediately problematizes this by pointing to operation cost, error rate, and downstream business cost.
itchy brain Emotional Impact of AI Work
Beck's name for the felt phenomenon: an idea or thought lodges in your head and the discomfort does not abate until you have made the thing in the world. Beck frames it as 'a bit more selfish' than Grinich's connect-with-people version; Grinich calls it 'a compulsion,' shared by long-distance founders as 'the same kind of mental disorder.' The durable version of the maker drive — opposed to external-validation chasing, which burns out.
picks and shovels
Beck's gold-rush framing for WorkOS's position: by selling infrastructure to everyone in the AI gold rush, Grinich gets to see what everybody is actually building, not just what gets press. Used to credential the across-the-ecosystem vantage claim.
enterprise ready Enterprise AI Adoption & Readiness
The bundle of capabilities a SaaS product needs to sell to large customers: SSO/SAML authentication, SCIM provisioning, access control, role-based permissions, audit logging. 'Not the fun stuff to build,' but required to grow up-market. The gap between 'running system' and 'mature fully running system.'
explore / expand / extract (3 X's) Explore / Expand / Extract
Beck's phase model: explore for product-market fit; expand as fast as possible once found; extract capital from what you've built to feed the next thing. Each phase has incompatible skill demands and tradeoffs. People retcon the expand phase into either inevitability or near-disaster.
industrialization phase (of AI) Industrialization of AI
Grinich's analogy: every culture going through industrialization burns coal, pollutes rivers, runs without baselines, then cleans up. We are in the dirty-transformation phase of AI — token budgets blown with no priors, money spent without optimization — and that is appropriate. Optimization comes after the baseline.
step function shift (vs. acceleration) Industrialization of AI
Grinich's claim that people are misreading the moment: they think AI is making things faster, when in fact the regime has changed. Acceleration is more of the same; step function means a different surface entirely. The 'goal for all these companies should be to get to the other side of the change as fast as you can.'
agency Agency
Grinich's framing — the realization (he places his own around age 18–19) that the world is malleable and that the people who change things 'just worked really, really, really hard.' He notes the term has become popular and uses it as the link between his geek interest (going deep) and his impact orientation.
geek Intrinsic Motivation & Making
Beck cites a definition (attributed to a writer whose name the auto-caption garbles as 'Jeep Hills' — likely Jerry Pournelle, Paul Graham, or similar; unverifiable here): someone who is highly creative, highly technical, and highly desirous of being both. Used to credential Grinich.
white-collar labor disruption AI Labor Disruption
Grinich's reading of what AI fundamentally is — not 'a big technology wave' like crypto or VR, but a labor disruption that will reshape how work and jobs function. He credits Amodei and agrees with him. The reason the change will be enormous and also why people are missing it.

Propositions

empirical (vendor-reported)WorkOS's vantage across hundreds of SaaS/AI customers shows that the whole ecosystem — not just AI labs — is accelerating: bigger TAM, faster growth, bigger companies adopting faster.conf: high
diagnostic / theoreticalThe current AI moment is a step-function regime change, not another acceleration; many people misread it as a hype wave because they have prior-wave fatigue (crypto, VR).conf: high
causal-empiricalLarge incumbents are adopting AI unusually fast — historically anomalous for them — because COVID built the muscle memory for compressed transformations ('five years of IT transformation in five weeks').conf: high
economic / structuralThe marginal cost of producing software has gone 'essentially to zero,' which will rebalance the labor demand for engineers — fewer engineers needed in any given company, more specialization, more years of training, but possibly more engineers overall because more software gets written.conf: high
structural / sociologicalcontestedSoftware engineering as practiced 2010–2025 (four-year undergrad → L3 at Meta → six figures) was historically anomalous; the profession will normalize toward medicine and law, with longer apprenticeship, specialization (e.g., a 'fellowship' in running 10,000 Postgres nodes), and smaller higher-paid in-house teams.conf: high
contradictsSupporting Your Employees' Career Growth When Everyone Is Overwhelmed · p5 — HBR says deep specialization is now an automation risk; Grinich predicts the profession normalizes toward medicine/law with more specialization and longer training.
contradictsDesigning the AI-native engineering organization · p8 — Beck argues the profession will normalize toward longer apprenticeship and deep specialization (medicine/law model), against the claim that generalism and product judgment beat deep technical specialization.
structural counter-trendThe barrier to getting software working is dropping at the same time as deeper expertise is gaining importance; less-skilled people will increasingly make decisions formerly reserved for the more skilled. Which trend dominates is unresolved.conf: high
methodological / normativeProfit is the honest measure of AI's impact on software, but most people use proxies (lines of code, PR count) because profit is hard to attribute.conf: high
competitive-dynamics / counter to BeckcontestedProfit cannot serve as the measure, because under Red Queen dynamics every competitor adopts the same technology, so the baseline shifts with you; speed (faster experimentation, faster market exploration) is the better indicator.conf: high
contradictsSimon Wardley's Post — Industrialisation of ML: nine notes on what to expect · p8 — Wardley says P&L impact is the right metric for AI-assisted development; Grinich argues Red Queen dynamics make profit unusable as a measure because every competitor's baseline shifts too.
empirical-organizational⚡ disputed & corroboratedThe role of the pure people-manager — the leader who delegates, defers, and creates psychological safety through process — is evaporating in favor of hands-on leaders driving customer value, product shipping, and business impact. CTOs are returning to code.conf: high
supports ←The engineering manager role is splitting in two · p3 — still-burning claims the pure people-manager is evaporating in favor of hands-on leaders who drive delivery; Moreau's TLM (senior engineer still writing code, setting technical direction) is exactly that hands-on future.
contradicts ←The engineering manager role is splitting in two · p4 — The same still-burning claim (pure people-manager evaporating) is contradicted by Moreau's second future, where the pure, non-technical people-manager does not evaporate but scales up to many teams — the article deliberately keeps both outcomes live.
structural concerncorroboratedIf engineering managers now have 50 direct reports, the historical 1:1 mentorship channel through which apprentices learned the craft from senior practitioners no longer scales; the apprenticeship layer is missing.conf: high
supports ←The engineering manager role is splitting in two · p4 — still-burning's conditional (if EMs have 50 direct reports, the 1:1 apprenticeship channel breaks) has its antecedent affirmed by Moreau, who presents the ~50:1 multi-team EM as a real emerging trajectory, not a hypothetical.
extends ←Update on Software Engineering Career Ladder · p3 — Both describe the same underlying failure mode -- a manager/mentor lacking the bandwidth or tenure-based depth of knowledge needed to do the people-intensive part of their role -- here from evaluation burden on new managers, there from headcount ratios breaking the 1:1 mentorship channel.
psychological / normativeThe durable motivation for founders and makers is the maker-compulsion itself — the 'itchy brain' / wanting to make and connect with people — not external validation, which 'cooks' you over time.conf: high
empirical (Beck's observation)Leadership has real leverage on team performance — a single addition or departure can transform a 10–30 person team's collective output while the other members stay the same.conf: high
practical-epistemicLate-stage company decisions (post product-market-fit) are made on 10–15% of the picture; waiting for 90% means you have already lost. Judgment under low information is the hardest leadership skill to learn.conf: high
historical-analogicalWe are in the industrialization phase of AI — token budgets blown by May with no priors, money spent without optimization — and that is appropriate; baselines and optimization come after, not before, the transformation.conf: medium
framing / normativeAt its core, AI is a labor disruption (especially white-collar), not just a technology wave; this is why its effects will be enormous and also why people are missing the magnitude.conf: high
practical-normative (closing metaphor)The correct posture is to paddle straight into the wave: people who try to sit it out or approach it sideways are the ones who get flipped.conf: high

Arguments

The whole software ecosystem (not just AI labs) is genuinely accelerating, and this is a regime change rather than a hype cycle. strength: medium

Explicit premises:

WorkOS sees revenue and growth patterns across hundreds of SaaS/AI companies and observes broad acceleration.
Code agents compress timelines (12-week projects to 20-minute runs).
AI models make products more human (natural-language interfaces beating SQL etc.), which alone would be a major technology shift.
Large incumbents are adopting faster than ever, in part because COVID built their transformation muscle.
Massive capital is flowing into compute, energy, and data centers — comparable historically only to railroads or wartime mobilization.

Implicit premises:

Each of these forces multiplies the others rather than substituting.
Vendor revenue is a reasonable proxy for ecosystem-wide adoption.

Inferences:

Multiple independent forces compounding implies step-function transformation, not linear improvement.
Comparable historical mobilizations (railroads, wartime) produced regime changes, not hype cycles.

Profit is the wrong measure of AI's impact; speed is the right measure. strength: high

Explicit premises:

Measuring AI's profit effect would require a counterfactual where only one company uses AI.
Every competitor is adopting AI (universal adoption).
By the Red Queen evolutionary logic, when everyone runs faster, no one outpaces the pack on absolute fitness — all the gains go to staying in the race.

Implicit premises:

Companies that fail to adopt drop out, so the comparison set self-selects.
Faster experimentation and exploration translate to harder-to-measure but real competitive advantages.

Inferences:

If profit gains are invisible in equilibrium, then the surviving signal is tempo: experiments per unit time, time-to-new-market.

Software engineering as a profession will normalize toward medicine, law, and traditional engineering — longer training, more specialization, smaller in-house teams, higher pay per specialist. strength: medium

Explicit premises:

Other knowledge-work professions universally require multi-year apprenticeship (residency, clerkship, architecture school's additional five years).
Software was uniquely a four-year undergrad → six-figure L3 pipeline, driven by enormous demand and value-creation.
With marginal code-production cost approaching zero, that demand profile changes.

Implicit premises:

Education systems and labor markets will route around the absence of in-the-job apprenticeship by adding pre-job training (the historical pattern in other professions).
The skills that survive are domain-specialized, judgment-heavy operation of large-scale systems — not generic coding.

Inferences:

If demand for generic coding falls and demand for judgment-and-specialization rises, the profession's structure must shift to match the other knowledge professions.

The maker-compulsion ('itchy brain') is the only durable founder motivation; external-validation motivations burn out. strength: high

Explicit premises:

Grinich's most miserable period was the 18 months between companies when he had no thing to chase.
Founders Grinich has observed who are driven by external validation 'get cooked' over time.
Beck names the same phenomenon ('itchy brain') from his own experience — the idea won't leave until you make the thing.

Implicit premises:

The Himalayan Sherpa's 'look at the step in front of you' generalizes: process-oriented motivation outlasts summit-oriented motivation.
Long-horizon endurance is what determines who is still building decades in.

Inferences:

If maker-compulsion is the durable substrate, then institutions/conditions that protect attention and the ability to make things are non-negotiable infrastructure.

Hands-on, customer-driving engineering leadership is replacing the pure people-manager. strength: medium

Explicit premises:

Across customers, Grinich sees hands-off delegation-style managers losing ground.
CTOs at big companies are returning to code or picking up AI coding tools.
Companies are reorganizing around 'driving forward customer usage and value' rather than around psychological-safety process.

Implicit premises:

What makes leaders effective in a fast-moving regime is direct contact with the value-producing surface, not orchestration of others doing it.
AI tools restore individual leverage to senior leaders, making hands-on contribution feasible again at scale.

Inferences:

When the technology floor of individual leverage rises, the optimal leadership style shifts from delegation to demonstration.

The correct posture toward the AI transformation is to head straight into it (paddle into the wave), not sideways or stationary. strength: high

Explicit premises:

The kayaking analogy: the stable position when a wave hits is bow-into-wave with active paddling.
The transformation is real (the step-function claim).
It will be a shock, but 'we'll make it through' — Grinich is constitutionally an optimist.

Implicit premises:

Inaction is not a neutral choice; the wave moves the ecosystem whether or not you engage.
Companies and individuals that engage early gain orientation that compounds.

Inferences:

Therefore: lean in; don't try to wait it out; the dangerous postures are paralysis and partial commitment.

Narratives

Questions addressed

⭕ Is the current AI boom genuinely different from prior tech cycles, or is it another crypto/VR-shaped hype wave?
⭕ What does 'the marginal cost of software approaches zero' actually entail?
⭕ How should we measure AI's impact on software?
⭕ What is happening to engineering management?
⭕ Where does mentorship/apprenticeship come from if managers have 50 reports?
⭕ What does Grinich fear?
⭕ What sustains a maker over a long career?

Gaps

incompleteVendor incentive distortion in the 'whole ecosystem is accelerating' claim
Read next: DORA 2025 and the Stanford/Microsoft developer-productivity studies on AI's actual code-quality effects; Gergely Orosz on enterprise sales-cycle compression; Marc Andreessen's classic 'why software is eating the world' for the prior pattern. For the contrarian counterweight: Gary Marcus and Ed Zitron on whether the AI productivity story is being narrated by the people selling it.
incompleteThe marginal-cost claim is accounting-incomplete (Beck flags it; Grinich doesn't return)
Read next: Anything in your Kent Beck file on TCR and scope-of-done (this is exactly the conversation about plausible deniability and shrinking scope); Lorin Hochstein on operational complexity; the Böckeler harness-engineering extraction in this repo (the 'gap between running system and trusted system' is the harness-engineering surface). The Adam Tooze / Brad DeLong line on 'productivity gains that don't show up in the statistics' is the macroeconomic complement.
incompleteThe apprenticeship transition is named but unanswered
Read next: John Cutler on overload, accountability, and DRI dynamics; Tanya Reilly 'The Staff Engineer's Path'; Will Larson on engineering organization growth. The persona-relevant cross-connection: the 'fragility of ideas' layer of Kjetil's reading — Fearless Change, Time to Think — applies directly. If mentorship is the institutional adoption channel for craft, its evaporation is the kind of organizational failure-to-receive-ideas Kjetil already reads for.
closed byShould You Still Become a Software Engineer in 2026? GitHub VP — Names a concrete transition mechanism (preceptor role, senior measured by seniors minted) for the apprenticeship transition that the prior episode left unanswered, though the cross-company free-rider problem remains unresolved.
incompleteThe 'descaling of software' counter-trend is named by Beck and left hanging
Read next: Steve Yegge on vibe coding; Gene Kim's vibe coding series; Simon Willison on the 'no-coder using LLMs to build real software' phenomenon; the Schoening agency-over-skills extraction in this repo (the agency thesis is the descaling-friendly side of this tug of war). Honest framing: Beck's craft instincts and Grinich's institutional instincts pull in opposite directions; you may have to choose.
incompleteGrinich's 'industrialization phase' analogy launders externalities
Read next: Tooze on energy transitions; the IEA's compute-electricity projections; the literature on whether the late-19th-century industrialization actually 'cleaned up' or just exported externalities. For Kjetil specifically: this is the systems-thinking lens — feedback loops, stocks/flows, who absorbs what — applied to the rhetoric of inevitability.
incompleteWhat changes about Beck specifically — the craft thinker in this regime
Read next: Beck's own recent Substack posts (Tidy First, the 'starving genies' piece, 'forest thinning'); your highlights on Beck for where he reframes craft for the agentic era; Hickey's 'Simple Made Easy' as the substrate craft argument; Ousterhout's complexity work. The persona-relevant cross-thread: this is where Kjetil's 'fragility of ideas' hidden layer touches the AI transition — if the craft of attentive code-reading is the protected attention surface, what protects it now?
incompleteWhat does the 'itchy brain' protect against, exactly?
Read next: Pressfield on Resistance; Cal Newport on deep work degradation under tool proliferation; the David Whyte/Maria Popova line on the difference between making and consuming; the persona's Bailey/Hyperfocus annotations on what attention needs in order to do anything. This is where the persona's hidden layer (fragility of ideas) and the episode's title concept (itchy brain) collide — and it is the most productive question the episode opens.

Closes gaps elsewhere

Designing the AI-native engineering organization · g1 — Directly addresses the avoided headcount question: near-zero marginal cost of software rebalances labor demand — fewer engineers per company, more specialization, but possibly more engineers overall because more software gets written (p4).
Designing the AI-native engineering organization · g4 — The gap asked for Kent Beck-style practitioners defending deeper technical craft against the generalism claim; Beck argues here (p5-p6) that specialization and longer apprenticeship gain importance as the barrier to working software drops.
Supporting Your Employees' Career Growth When Everyone Is Overwhelmed · g3 — Supplies the requested pushback on the breadth-hedges-automation thesis: Grinich argues the profession normalizes toward medicine/law-style deep specialization with longer apprenticeships, while flagging the opposing barrier-dropping trend as unresolved.

Itchy Brain (Still Burning: Michael Grinich & Kent Beck on AI adoption, enterprise developer tools)

Structure

Key terms

Propositions

Arguments

Narratives

Questions addressed

Gaps

Closes gaps elsewhere

Related documents