You Don't Get to Create Anything (Still Burning: Beck + Randy Shoup on Jevons paradox and distributed systems under AI)

Kent Beck — Still Burning Podcast · podcast · 2026-06-03 Reader 🖥️

AI Labor Disruption Craft Community The Genie (Beck)Golden Path Guardrails Intrinsic Motivation & Making Jevons Paradox Near-Zero-Cost Code

Cognition is becoming near-free, and per Jevons paradox that will produce more cognitive work and more engineers, not fewer — but the people best prepared for this regime are exactly the ones who already lived through writing the distributed-systems playbook, because they know two things the panicking crowd forgets: that a playbook is hard-won and not obvious until written, and that the human compulsion underneath all of this is the maker's need to create the thing rather than merely transcribe it (the 'you don't get to create anything' moment that drives Shoup out of patent law and back into engineering); accordingly the right posture is calm, perspective-keeping, deductive engagement with the new regime — bounding the genie with specs/evals, making good practices easier rather than shaming the bad, and treating this as another wave of physics-discovery rather than a hype cycle or an apocalypse.

mixed · mixed — A long-form podcast conversation between two senior craft thinkers — Beck (host, software design tradition) and Shoup (guest, distributed-systems veteran of eBay/Google/Stitch Fix/WeWork/Thrive). Practical in its diagnoses (engineering management under AI, how to bound the genie, evals + spec-driven development) and theoretical in its framing (Jevons paradox on cognition, the playbook-writer/playbook-applier distinction, the deductive vs. inductive thinker split, the climate-policy aside about making good behavior easier rather than shaming). Form is mixed: narrative (Sand Hill Road summer-associate identity crisis, 2006 eBay-architecture talk at Cubberley, Shoup's geek-father lineage), expository (Shoup's mid-career biography: Oracle → Stanford Law → back to Oracle → eBay), argumentative (the playbook-was-not-obvious case, the Jevons-on-cognition forecast). No single explicit thesis; the joint thesis is constructed through Beck's questions and Shoup's reflections.

Structure

Part	Role
Opening sponsor frames (WorkOS + Augment Code) — not part of the conversation	Beck-as-host reads the sponsor copy: the gap from idea to running system has narrowed; what remains is purity, auditability, identity (CISO list). Augment positions itself as 'beyond the IDE': the genie does detail work while humans orient and decide. These are ad-reads, not Beck's spontaneous claims, but they set the frame the conversation lives inside.
Shoup's geek-origin: father at Xerox PARC, double-major at Stanford, internship at Intel	Establishes that Shoup grew up inside the computing tradition (PARC father) but resisted it as identity. Political science and international relations were the 'mainline'; CS/math was 'side gig, for fun.' Studied in West Berlin (wall still up) and Krakow (behind the Iron Curtain).
Sand Hill Road summer-associate: 'you don't get to create anything'	The autobiographical centerpiece and the source of the episode title. Shoup is on track for the four-year JD/MBA + international-relations program; the summer between Stanford Law and the IR school he takes a summer associateship doing IP / patent prosecution on Sand Hill Road. He sits with inventors at whiteboards, vibrates with excitement, wants to engage — and is told, repeatedly: no, you don't do that, you just write it down. Beck supplies the line: 'you don't get to create anything.' Shoup confirms it: 'all you do is write down the cool creation that the inventor did.' Identity crisis; he reads M. Mitchell Waldrop's Complexity as antidote; returns to Oracle as a software engineer.
International-relations background as appreciation for nuance	Beck asks whether the poli-sci/IR background plays into Shoup's executive work. Shoup says no counterfactual, but: social sciences cultivate appreciation for nuance, for the gray-ness of most of the world, for there being multiple ways to achieve a goal collectively. Technical disciplines have right/wrong at the bits-and-bytes level but not at the human level.
Perspective on the current moment: serious without panicking	Beck reads history; Shoup studied authoritarianism (says he wishes he hadn't taken those classes given the current climate). The shared posture: 'we have a disaster on our hands. No — this is not a disaster.' Any number of really horrible 20th-century things make this not even a bump. You can take a situation seriously without panicking. This is the meta-disposition of the entire conversation.
Why Shoup speaks publicly: 43 years of public speaking + eBay 2006	Two answers. First: he started high-school public speaking 43 years ago; he gets nervous about content (impostor: 'this is the year they find me out') but not about speaking. He enjoys the performative aspect. Second: in 2006 he and Dan Pritchett — the eBay architecture team — fought to give the first eBay-internals talk. Standing room only at Cubberley Auditorium in Palo Alto; got invited to inaugural QCon (Pritchett went). That started Shoup on conference speaking permanently.
Discovering the physics: eBay, Google, Yahoo, Amazon all hitting the same wall in parallel	Shoup's framing: 'there's some physics to scale, and once you discover the physics, they're the physics.' Tens to hundreds of millions of users; racks upon racks; humanity had never seen these problems. eBay, Yahoo, Amazon, Google were co-discovering the same techniques in parallel. Big companies were tight-lipped (Amazon, Google, eBay); Yahoo and — after Shoup and Pritchett won the fight — eBay began talking.
Why companies should let engineers speak: thinking tool + recruiting + training	Beck asks: what was in it for eBay? Shoup: in retrospect, it was a recruiting tool ('makes them look hot and cool'). But more importantly: 'formulating your thoughts in a way that is understandable by an external audience that doesn't work for you is a fantastic thinking tool... no better way to crystallize your own thinking than to try to explain it concisely and clearly to another person' (cites Einstein and Feynman). And: his old talks were still being used as internal training when he returned years later.
The playbook claim: people are used to turning to a page and applying the technique	Beck names the conversation's pivot: for years now there's been a playbook. Shoup confirms: people have gotten used to turning to a page and applying the technique. Now augmented development has 'wiped a lot of the playbook clean,' and people who have never not had a playbook are panicked. They don't realize: it's all right, you can write the next one — but that's a different set of skills.
Playbook-writing vs. playbook-applying are both hard — and the playbook is not obvious	Beck asks if Shoup got bored once the playbook existed. Shoup: no. Both are challenging in different ways. He praises Extreme Programming Explained (1999) as 'a new book that teaches you how to do AI' (his joke on LinkedIn); the trajectory continues through continuous delivery, Accelerate, frictionless. He pushes back at Beck warmly: 'the reason why you had to write that playbook is because it's not obvious. It is not obvious to most people that you shouldn't batch all your stuff up and try really hard on it... why would I write the test of a thing before I write the thing? That's crazy pants.' The playbook's invisibility-once-internalized is the whole challenge of transmission.
Beck's split: gets bored once the playbook exists; deductive Shoup loves applying it	Beck admits he gets bored once a playbook is written; he's ready to move on. Shoup names himself a deductive thinker (give him principles, he applies them) where Beck is inductive (sees examples, derives principles). They name the yin/yang: Beck's ideas need executors; without executors the ideas don't dent the world. Without ideas the executors do nothing interesting.
Augmented development at Thrive: dojos, knowledge graph, evals harness	Shoup was SVP of Engineering at Thrive Online Grocery (~100 engineers); just hired Mike Winslow to lead AI augmentation across the whole company. Dojos pair non-engineer humans with engineer humans to build things (the legal team, the merchandising team — 'everybody's super excited'). For AI augmentation of software specifically: 'it's all about the context and bounding the genie.' Two directions: spec-driven (forward) and eval / adversarial (backward-pushing). Plus a knowledge-graph project ('genome') that captures what 12-year-old software actually does as context for spec work and evals.
Shoup at home is a Luddite	Beck asks if he falls into the side-project trap (so easy to start things now). Shoup: not personally; he's 'a Luddite in my personal life' — came late to cell phones, doesn't run agents at home, no judgment of those who do. Loves the industry-scale distributed work at work; hikes, farmers' market, cooking, his son at home. Cooking is 'just chemistry' — still loves it. Has tried to get excited about technical side projects and can't.
Tools he wishes existed: visualization of organizational bottlenecks	Beck asks: what tool would you want? Shoup names something Mike Winslow just built — a dashboard showing team-by-team execution bottlenecks (feature rate, bug-fix rate). He's a visual thinker; can hold a lot of abstraction in his head but can't communicate it without visualization. He'd want 'show me the problems, in a visual way' across an entire engineering org. Beck: that's a business model for someone to go make.
Limit-WIP returns: Martin Thompson's 'when you show me you can use one thread, I'll give you another'	Beck notes it's easy to start things now but just as expensive to maintain. The admonition to limit WIP is one he's been repeating his whole career and still has to repeat. Shoup quotes Martin Thompson: 'when you show me you can use one thread, I'll give you another.'
What's next: CircleCI interim role, possibly overseas eventually	Shoup's Thrive role is ending. Almost certainly an interim role at CircleCI to help his network there (he's been coaching their CTO Rob Zuber and principal engineers for years). They are 'working hard to try to see what is the new agentic augmented team structure, project structure' — not just for customers but for themselves. Rob Zuber is leading a tiny two-three person team going at 'blazing light speed.' Longer term: Shoup has always wanted to work overseas (West Berlin and Poland college years; never had the chance). Son is at Yale doing mechanical engineering, so empty-nest mode unlocks possibilities.
Jevons paradox on cognition: the central forecast	Shoup names it directly: 'this is Jevons paradox now that essentially cognition is near free... now there are all these things that we would do with cognition that were not economically viable and now they are.' Walks through the original Jevons case (1850s–1860s English economist; coal extraction got 10x cheaper; net spend on coal went up because new uses opened). Applies to AI: he is 'confident in the long run per Jevons paradox that it will create more jobs and there will still be not just the same number of engineers but more.' Pairs it with the agricultural analogy: in 1900, 80% of US population was in agriculture; now 2–3%. If you'd told a 1900 farmer that 78% wouldn't be farming in 126 years, they'd have called it dystopia. Wouldn't have been 100% wrong — but the transformation also created enormous new things they couldn't imagine.
Closing political-scientist coda: make good behavior easier, don't shame	Beck asks what scares Shoup. Shoup gives a small set ('not the top 150'). Cautiously optimistic about country, society, planet — every one of them under threat in new ways. The only way forward is through, and through means using technology. Then the operative principle, delivered as the technologist and political scientist together: 'we have learned over and over again that you can't shame people into doing right stuff.' Making people feel bad about hurting the environment is not going to make them save the environment. Instead, make it easy and good and objectively better for everybody. Many people who share his political beliefs don't fully see that the economic forces you can really harness are 'I need life to be better for me and my kids' — not the morality of self-deprivation.

Key terms

you don't get to create anything Intrinsic Motivation & Making
Beck supplies the phrase as the operational definition of the scribe role: the patent prosecutor watches the inventor on the whiteboard and is forbidden to engage — 'all you do is write down the cool creation that the inventor did.' For Shoup, hearing it crystallizes his identity crisis: he is constitutionally a creator (and his career proves it). In the show-notes framing it carries a second resonance: the implicit fear about AI is that engineers become scribes for the genie — transcribing what the model invents. The Sand Hill Road story argues against that frame: people who need to create will refuse the scribe role, and the playbook-writers of past regimes were exactly those refusers.
Jevons paradox (on cognition) Jevons Paradox
Shoup's central forecast device. Original (1850s–1860s, William Stanley Jevons): when coal extraction became 10x cheaper, net spending on coal rose, because at the lower price entirely new uses became economically viable. Shoup's application: cognition is now near-free; therefore tasks that were not economically viable as cognitive work will now be done, and the number of engineers (and jobs) will rise, not fall. He pairs it explicitly with the 1900-agriculture analogy to argue that catastrophic-sounding workforce shifts can be true at the surface and benign-to-positive in long-run aggregate.
the playbook Craft Community
Beck's word for the accumulated craft of software — TDD, small steps, continuous integration, the distributed-systems patterns, microservices, observability — each codified into pages that practitioners can turn to. Shoup confirms it as 'the physics' that gets discovered once and then is just the physics. The conversation's diagnosis: AI has 'wiped a lot of the playbook clean,' and a generation of engineers who have never not had a playbook are panicking because they don't know they're allowed to write the next one. Crucially: writing a playbook and applying one are both hard, but they require different skills.
bounding the genie Guardrails
Shoup's phrasing for the AI-augmentation discipline he's running at Thrive: 'it's all about the context and bounding the genie — how can we keep the genie within the bounds we would like.' Two complementary directions: spec-driven development (forward: the genie is told what to make) and evals + adversarial testing (backward-pushing: the genie's output is constrained by what passes). Picks up Beck's prior 'starving genies' framing without naming it directly.
the physics (of scale, of software, of cognition)
Shoup's metaphor for the underlying truths of a domain that exist independently of who discovers them. 'There's some physics to scale to scaling and once you discover the physics, they're the physics... we were all discovering the physics all at the same time.' The eBay/Yahoo/Amazon/Google parallel discovery proves the metaphor: independent groups converged on the same techniques because the underlying constraints were the same. The implicit extension: AI under cognition-is-cheap has its own physics, currently being discovered in parallel, and writing it down is the craft.
deductive vs. inductive thinker
Shoup names the difference between himself and Beck. Shoup is deductive: state the principles, then apply them — 'give me a platonic ideal or whatever and then I can apply it in the real world.' Beck is inductive: see examples, derive the principle, get bored once the principle exists, move on. Both temperaments are necessary; the conversation explicitly frames them as yin/yang.
starving genies / bounding the genie / genome
Three related terms in the conversation's vicinity. 'Bounding the genie' is Shoup's (above). 'Genome' is the codename of the Thrive knowledge-graph project: a collective description of what the 12-year-old product actually does, intended as context for spec work and evals. The Beck-coined 'starving genies' is not named explicitly but the conceptual proximity is unmistakable — the conversation lives inside Beck's reframe of LLMs as genies that must be fed bounded context.
the descaling / industrialization of cognition Near-Zero-Cost Code
Not named explicitly but implicit in the Jevons + 1900-agriculture pair: the comparative-advantage shift that hollowed out a category of labor (farming) while creating categories the displaced couldn't imagine. Shoup's forecast is that cognitive work will follow the same arc — not the same shape, but the same logic.
make it easy / don't shame Golden Path
Shoup's closing meta-principle, delivered as the technologist + political-scientist together. People do not change behavior through shame; they change behavior when the better path is objectively better for them and their kids. Stated about climate, but immediately generalizable to AI adoption and organizational craft transmission. The principle implies that rolling out new playbooks (TDD, evals, spec-driven, distributed-systems patterns) succeeds through making the path of least resistance the right one.
perspective / 'not even a bump'
Beck and Shoup share a disposition: take the situation seriously without panicking. Beck reads history; Shoup studied authoritarianism. 'Any number of really horrible things that have happened in the last hundred years that make this not even a bump.' This is the meta-mood of the entire conversation — the precondition for the playbook-writing temperament rather than the playbook-panic temperament.

Propositions

economic forecast / framing⚡ disputed & corroboratedCognition has become near-free, and per Jevons paradox this will create more cognitive work and more engineers, not fewer — because tasks that were not economically viable as cognitive work now are.conf: high
supports ←Presentation: The Free-Lunch Guide to Idea Circularity · p7 — Same Jevons Paradox claim applied to cognitive/software labor -- cheaper cognition/code production increases total demand rather than reducing headcount.
contradicts ←AI and the Looming Competition for Margin · p5 — This article's margin-erosion mechanism for 'more with same' explicitly requires inelastic demand; the corpus's Jevons-paradox claim that cheaper cognition creates MORE cognitive work (elastic demand) implies that, by this article's own logic, cognitive/software industries should be relatively exempt from the margin erosion it treats as broadly applicable.
supports ←Research: How AI Agents Broaden the Scope of Knowledge Work · p9 — Beck/Shoup predict from Jevons paradox that near-free cognition creates more cognitive work by making previously-uneconomic tasks viable; this piece's field data — 23-40% of agent tasks were never attempted with the assistant, plus agent adoption increasing (not cannibalizing) assistant usage — is a concrete empirical instance of that mechanism.
historical-analogical forecastThe transition will produce major disruption of the humans currently employed in cognitive work — analogous to the agricultural shift from 80% of the US workforce in 1900 to 2–3% today — but the long-run transformation creates entirely new categories of work the displaced cannot foresee.conf: high
refinesSimon Wardley's Post — Industrialisation of ML: nine notes on what to expect · p4 — Accepts Wardley's no-jobs-apocalypse frame but adds the qualification that the transition still inflicts major displacement on current cognitive workers (agriculture analogy).
diagnosticThere is a playbook for distributed-systems work at scale (and for software more broadly) that took a generation to write; people have gotten used to turning to a page and applying the technique; the AI moment has 'wiped a lot of the playbook clean.'conf: high
diagnostic / sociologicalThe people panicking about AI are mostly playbook-appliers who have never not had a playbook; they don't realize they're allowed to write the next one. The playbook-writers (those who lived through writing the original distributed-systems playbook) aren't panicking, because they know what writing one feels like.conf: high
structural / craftWriting a playbook and applying one are both hard, but they require different skills. Both are creative; both require working with the physics of the domain; neither is the residual of the other.conf: high
epistemic / theoreticalThere is a physics to scale (and by implicit extension, to AI under cognition-is-cheap). Once you discover the physics, they're the physics — multiple independent teams (eBay, Yahoo, Amazon, Google in 2006) converged on the same scalability techniques because the underlying constraints were the same.conf: high
epistemic / methodologicalPublic explanation is a fantastic thinking tool — there is no better way to crystallize one's own thinking than to try to explain it concisely and clearly to another person who doesn't work for you (Einstein/Feynman lineage). Conference talks are how senior practitioners reproduce the craft.conf: high
psychological / autobiographicalThe maker drive — the need to create the thing rather than transcribe it — is durable enough to overturn ten years of career planning (Shoup walks out of patent law mid-summer after the 'you don't get to create anything' realization).conf: high
practical / craftAI augmentation of software work should proceed through bounding the genie: spec-driven development (forward) plus evals and adversarial testing (backward), with a knowledge graph of what existing software actually does as context.conf: high
extendsThe Future Of Software Engineering · p2 — Generalizes TDD-as-bounding into a two-directional regime: spec-driven development forward, evals and adversarial testing backward, with a knowledge graph as context.
supportsAgents don't know what good looks like. And that's exactly the problem. · p5 — Bounding-the-genie (spec-driven forward plus evals and adversarial testing backward) is a concrete restatement of deterministic guardrails around non-deterministic agents.
structural / craftWhen cognition is cheap, the durable constraint shifts to attention and execution capacity (work-in-progress). Limit-WIP — 'when you show me you can use one thread, I'll give you another' — remains the discipline.conf: high
structural / psychologicalDeductive and inductive temperaments are complementary, not interchangeable. Ideas (inductive) need executors (deductive); without executors the ideas don't dent the world; without ideas the executors do nothing interesting.conf: high
dispositionalThe right meta-disposition toward the current moment is to take it seriously without panicking. History and perspective — reading the 20th century, studying authoritarianism, knowing that worse has happened — license the playbook-writing temperament instead of the playbook-panic temperament.conf: high
normative / change-managementInstitutional change happens by making the better path easier, not by shaming the worse. People will not adopt the new playbook because they are told they should; they will adopt it because it is objectively better for them and their kids.conf: high
psychological / craftDurable engagement with one's craft does not require maximalist consumption of it outside work. A Luddite-at-home posture is compatible with bleeding-edge industrial work and may even be required for it.conf: high

Arguments

More cognitive work and more engineers, not fewer, in the long run — even as the current cohort experiences major disruption. strength: high

Explicit premises:

Cognition has become near-free with AI (the underlying price-of-cognition shift).
Jevons paradox: when an input's effective price falls, total demand for the activities that consume that input rises, often dramatically.
The agricultural analogy: 1900 had 80% of US population in agriculture; 2026 has 2–3%, and the displacement created entirely new categories of work.

Implicit premises:

Cognition is similar enough to coal/farming as an input that the rebound mechanism applies.
New categories of cognitive work exist below the current price floor and will appear as the price drops.
Major disruption of currently-employed humans is a real cost but does not invalidate the long-run trajectory.

Inferences:

Cheaper cognition opens uses that were previously uneconomic.
Those uses absorb more cognitive labor than the displacement releases.
Therefore total cognitive work — and the engineers who do it — grows.

The right response to AI is to write the next playbook — not to panic, not to apply the old one, not to wait for someone else to write it. strength: high

Explicit premises:

There has been a playbook (distributed-systems patterns, TDD, continuous delivery, the trajectory through Accelerate).
AI has wiped a lot of the playbook clean.
Writing a playbook requires the skill of seeing what isn't obvious; applying one requires the skill of execution.
The original playbook was discovered as physics — multiple teams converged independently because the underlying constraints were the same.
The people who wrote the original playbook know that writing one is hard but possible.

Implicit premises:

AI under cognition-is-cheap has its own physics, currently being discovered in parallel by multiple teams.
The first-generation playbook-writers are well-positioned to write the next one because they have the temperament.
Panic comes from never having seen a playbook get written.

Inferences:

Therefore: identify the physics; codify it; transmit it through teaching (conferences, dojos, training).

AI augmentation of software work succeeds through bounding the genie with spec-driven development and evals. strength: medium

Explicit premises:

Context is everything: the genie's behavior is shaped by what it knows about the target.
Spec-driven forward direction tells the genie what to make.
Evals and adversarial testing in the backward direction constrain the genie's output to what passes.
A knowledge graph of existing software behavior (the 'genome' project at Thrive) provides the context for both directions.

Implicit premises:

The genie's output is too unconstrained to be useful by itself.
Existing codebases (12-year-old systems) embody behavioral knowledge that is not encoded anywhere readable by the genie until explicitly captured.
The same discipline that wrote the test-first / continuous-delivery playbook generalizes to AI augmentation: small steps, attentive constraints, fast feedback.

Inferences:

Therefore: invest in spec discipline, evals harnesses, and knowledge graphs as the augmentation infrastructure.

Institutions adopt new craft through making the better path easier, not through shaming the worse. strength: high

Explicit premises:

Climate-policy evidence: shaming people into environmental virtue has not worked at scale.
Behavioral economics: people change behavior when the alternative is objectively better for them and their kids.
The only economic forces you can harness are 'life is better for me and my kids.'
Moral self-deprivation is not how most populations behave, even when small minorities have that morality.

Implicit premises:

Software craft adoption follows the same pattern as climate-behavior adoption (it is a behavior-change problem at root).
Therefore advocacy for TDD, spec-driven, evals, distributed-systems patterns, etc., must reduce friction rather than increase guilt.

Inferences:

Therefore: design organizational rollouts of new playbooks so that the right thing is also the easy thing.

The maker drive is what survives every regime change — the people who refuse the scribe role are the ones who write the next playbook. strength: high

Explicit premises:

Shoup's Sand Hill Road experience: forced into the transcriber role, he could not survive it.
Ten years of identity around international law collapsed in a single summer because the role denied him the chance to create.
Beck names the same pattern as 'creative in the true sense of creating a thing.'

Implicit premises:

The implicit fear about AI is exactly the scribe role: engineer-as-transcriber for what the model invents.
People with the maker compulsion will refuse that role regardless of the financial or institutional pressure to accept it.
Those refusers are the ones from whom the next playbook is written.

Inferences:

Therefore: the maker compulsion is structural, not optional; the regime that survives is the one that channels it rather than represses it.

Narratives

Questions addressed

⭕ What does it mean to be a creator, and what does it mean to be denied the role?
⭕ What happens to engineering when cognition becomes near-free?
⭕ Why aren't the people who wrote the original distributed-systems playbook panicking about AI?
⭕ How should engineering teams approach AI augmentation in practice?
⭕ How should institutions roll out new craft — whether climate policy, AI, or software discipline?
⭕ What posture should the senior practitioner take toward the current AI moment?
⭕ What constraints survive when cognition is cheap?
⭕ Do you need to be maximally engaged with your craft to stay sharp?

Gaps

incompleteThe 'you don't get to create anything' extension to AI is never made explicit
Read next: Beck's own Substack on 'starving genies' and the human's relationship to the genie's outputs; Cal Newport's deep-work tradition on how the substitution of tool-mediated production changes the nature of the work; Pressfield on Resistance; and — the persona-relevant cross-thread — the 'fragility of ideas' hidden layer in Kjetil's reading: if the maker-compulsion is the durable substrate, what does it look like when the agent satisfies it without you having moved? This is the productive question the episode leaves on the table.
incompleteThe Jevons-on-cognition forecast is asserted, not argued through to its qualitative differences
Read next: Carl Benedikt Frey 'The Technology Trap' on the political economy of transitional displacement; Brad DeLong on what 'creating new categories of work' has actually looked like historically; Anton Korinek and Daron Acemoglu on AI-and-labor specifically. For the contrarian side: Erik Brynjolfsson on the J-curve (gains that show up late and unevenly); the Robin Hanson 'Age of Em' tradition for the qualitative-difference-of-cognition objection. Persona-relevant: DORA 2025 on AI's actual code-quality effects.
incompleteThe playbook claim under-specifies the AI playbook's structure
Read next: Hamel Husain on evals; Simon Willison on the working LLM-augmented practitioner's actual playbook; Addy Osmani and the Claude Code patterns crowd; the Bernhard / Bockeler / Thoughtworks harness-engineering thread already in this repo; the Heis ep 11 AI-special with Vilde Opsal and Anders Sveen for the Norwegian practitioner cut; the Heis ep 13 Torrissen for code-quality-under-AI. This is exactly where 'craft transmission for the next era' lives — and where the conversation is most incomplete.
incompleteThe 'apprenticeship channel under 50:1 manager ratios' problem from the prior Still Burning episode is not picked up
Read next: John Cutler on overload and DRI dynamics; Will Larson and Tanya Reilly on engineering-org structure; the Heis ep 13 Torrissen extraction for the Norwegian practitioner take on code-quality and craft transmission under AI. Persona-relevant: the 'fragility of ideas' layer — ideas die at the institutional adoption point if the apprenticeship channel is gone.
incompleteShoup's 'make it easy, don't shame' principle is stated as climate policy and not applied to software craft
Read next: Cutler on operating-system design and waste; Thaler and Sunstein on choice architecture for the foundational behavioral-economics text; Kniberg on Spotify's model as an example of friction-reduction at scale. Persona-relevant: this is exactly where Kjetil's leadership-vs-management orientation and his Cutler-deep reading already live.
incompleteThe 'distributed-systems-playbook-writers as exemplars' framing implicitly leaves out the playbook that didn't get written
Read next: Schneier on security (in Kjetil's RSS); the Ruined by Design / Mike Monteiro line on what professionals fail to write playbooks for; the Norwegian intelligence reports (NSM, PST) for the security-and-resilience perspective; recent regulatory writing on AI-and-governance for what is being written by non-engineers because engineers didn't write it.
incompleteThe Beck-as-inductive vs. Shoup-as-deductive split is named but not connected to the playbook-writing claim
Read next: Anything on creative-cognitive partnerships at scale: Brian Eno's 'scenius' concept, Anne Lamott on the writer-editor pairing, the literature on the pair-programming and mob-programming traditions in software. For the temperament-and-craft thread specifically: Beck's own writing on the craft of seeing simple design where others see complexity.

You Don't Get to Create Anything (Still Burning: Beck + Randy Shoup on Jevons paradox and distributed systems under AI)

Structure

Key terms

Propositions

Arguments

Narratives

Questions addressed

Gaps

Related documents