Taste & Judgment

The human capacity to judge what is good — design, architecture, output quality — with no objectively checkable answer; the faculty AI cannot substitute and agents amplify.

Also known as: taste, product taste, judgment, smak, obviously good, hypothesis quality

How authors use it

Oldest first (extraction date when no publication date is known).

Documents (32)

• AI + Teams = ? Marcus Castenfors · 2026-09-26 source ↗ Reader 🖥️
  The default way people use generative AI on a team — everyone working alone in their own chat window — quietly damages team alignment and accrues a 'collective debt'; teams do better when they deliberately decide where to prompt together (especially to define the problem, which is a human act AI cannot do well) versus where to delegate to AI individually or collectively, and when they widen the room to outside domain experts now that AI has removed the technical barrier to those experts contributing directly.
• AI Is Changing the Rules of Entrepreneurship Victor P. Seidel, Bret Greenstein, Thomas H. Davenport · 2026-09-09 source ↗ Reader 🖥️
  AI has moved entrepreneurship from a resource-scarcity paradigm (the lean startup, optimized around scarce capital and labor) to a resource-abundance paradigm the authors call 'abundance entrepreneurship,' in which a single founder can orchestrate AI agents to generate, build, and run many ventures in parallel -- so the challenge for entrepreneurs and established firms alike is no longer gaining access to scarce resources but organizing abundant ones, which shifts the binding constraint to founder attention, judgment, and the domain expertise needed to tell which abundant option is actually worth pursuing.
• Own the Outer Loop Addy Osmani · 2026-09-09 source ↗ Reader 🖥️
  Because agents supply capability but cannot inherit consequence, engineers must own the 'outer loop' of agentic systems — the boundary where Quality (checks/evidence), Verdict (the ship/block decision), and Answerability (the guarantee of being able to explain why) live — which means personal, attributable accountability, not code-writing skill, is what becomes scarce and valuable as harnesses, loops, fleets, and software factories absorb the inner work.
• How Anthropic Builds And How Engineering Will Change Soon | Thariq Shihipar Ryan Peterman · 2026-09-07 Reader 🖥️
  Inside Anthropic, engineering is already organized around treating Claude as a thought partner wrapped in an increasingly sophisticated harness/loop rather than around hand-writing code, and this shift will diffuse into the rest of the industry — leaving the durable human value in judgment, taste, verification design, and self-directed visibility rather than in code production itself.
• How to turn your AI into a world-class designer Anshu Chimala · 2026-09-01 Reader 🖥️
  AI models default to bland, repetitive ('design-by-committee') design because their training rewards the statistically safest choice at every token, so producing distinctive AI-generated design requires deliberately counteracting that bias at each stage of a Discover/Define/Deliver process — injecting externally-sourced variety and ambition, judging output with a fresh-context critic instead of the generating agent, enriching it with image/video generation, and then subtracting rather than adding to reach a polished result.
• Presentation: Getting Rid of LeetCode Interviews in the World of AI Daniel Doubrovkine · 2026-07-29 source ↗ Reader 🖥️
  Because algorithmic whiteboard interviews measure luck and recent rehearsal rather than genuine engineering ability — as the speaker's own repeatedly inconsistent interview record proves — companies should replace them with open-ended, AI-assisted system-design exercises that reveal the judgment and human skills AI cannot substitute.
• Why Mentoring Matters More in the AI Era Marlo Lyons · 2026-07-24 Reader 🖥️
  Because AI is automating the routine work that once built professional judgment and pattern recognition through repetition, organizations must replace that lost experiential foundation with deliberately engineered, intentional mentoring — not the ad hoc, exposure-dependent mentoring models that assumed hands-on learning was already happening.
• Addy Osmani’s Post Addy Osmani · 2026-07-11 source ↗ Reader 🖥️
  As coding agents make solving cheap and abundant, an engineer's durable career value shifts from being the fastest problem-solver to the 'ungradeable' meta-skills that have no answer key — selecting which problems matter, judging quality through earned taste, directing agents rather than doing the work, sprinting the hard last mile, and owning the outcome.
• Adam Mosseri: AI is a tailwind for authenticity Adam Mosseri (guest); Lenny Rachitsky (host) · 2026-07-11 Reader 🖥️
  Because AI makes creation nearly free and floods the world with synthetic content, the scarce and durable human advantages become taste, judgment and authenticity — which Mosseri argues is a tailwind rather than a headwind for Instagram (audiences will seek out real people and creators, not less), and which in turn reshapes how work is organized: teams compress into small generalist 'pods' led by a new 'product staff' role, while hiring rewards curiosity, grit and a willingness to try over deep specialization.
• Det skal være vanskelig Erik Børve Rasmussen · 2026-07-10 source ↗ Reader 🖥️
  Difficulty is the mechanism of learning, not a defect in it, so AI — a machine for removing difficulty — destroys learning wherever the process rather than the product was the whole point (school, unlike work); the remedy is therefore to redesign AI and assessment as skill-forcing 'simulators' rather than to ban them.
• Podcast: Formal Methods for Every Engineer in an AI-Powered Future Gabriela Moreira (guest); host Shane Hastie (InfoQ) · 2026-07-10 Reader 🖥️
  Formal methods are a practical tool for reasoning about the edge cases of complex distributed and concurrent systems, and because AI in 2026 collapses the historic barriers to writing specifications and the glue code that binds them to real implementations, the profession's enduring human work shifts to defining which system behaviours are correct — the judgment AI cannot take over.
• Vibe coding has escaped the terminal Casey Newton · 2026-07-08 source ↗ Reader 🖥️
  Vibe-coding tools have escaped the terminal and IDE into consumer-friendly, GUI-native app builders like Raycast's Glaze, making it cheap and fun enough for a non-technical user to build genuinely bespoke, hyper-personalized desktop software — which gives Newton modest but real confidence that such software has a near-term future.
• Notes from FOSE Europe Giles Edwards-Alexander · 2026-07-07 source ↗ Reader 🖥️
  Five months into the agentic era the value of agentic engineering is now settled ('the evidence is in'), so the live frontier of the debate has moved from the tools themselves to the socio-organisational questions they raise — how to design your organisation (Optimiser vs Learner), where engineering rigour and human judgement now live, and how to coordinate agentic work across teams and enterprises — none of which is yet resolved.
• We've Stopped Arguing About Whether AI Changes Software Greg Herlein · 2026-07-07 source ↗ Reader 🖥️
  The debate over whether agentic AI transforms software engineering is settled; the transformation is a shift in the center of gravity of the work — from writing code to specifying, verifying, and curating knowledge.
• The hidden pattern behind successful products | Mark Pincus (Founder of Zynga) Lenny's Newsletter · 2026-07-06 source ↗ Reader 🖥️
  Successful products come not from bold original visions but from disciplined humility: trust your gut instinct while distrusting your specific ideas, master and tastefully copy what is already proven before layering on one unanimously-endorsed 'better' and one expendable 'new', start embarrassingly small, and act on belief grounded in real signal rather than on hope.
• SWLW #710: Revised rules of Eng Leadership, When AI costs more than the engineer, and more. Oren Ellenbogen · 2026-07-06 Reader 🖥️
  As AI reshapes engineering, the decisive leadership work shifts to human judgment applied to fundamentals — reaffirming sound operating principles, justifying AI's now salary-scale cost, and demanding engineering rigor — because AI only amplifies (in magnitude, not direction) whatever practices, clarity, and dysfunction already exist.
• Why "good enough" is the whole point of 80/20 Ernie Svenson · 2026-07-03 source ↗ Reader 🖥️
  The real value of the 80/20 rule is not the input-output ratio but the permission it grants to stop at 'good enough' — the outcome-relative level a job actually demands — because the last increment of effort usually goes unnoticed, so the true discipline is the confidence to quit polishing rather than the diligence to do more.
• You can't unit test for taste Karl Tryggvason · 2026-06-22 source ↗ Reader 🖥️
  In software you can assert correctness with red/green tests, but taste — judging what output is good, relevant and well-proportioned where there is no ground truth — cannot be unit-tested, so it stays an irreducibly human, iterative judgment even when AI does the production work.
• The best response to AI slop, infinite advice, and online noise is from Robin Williams Jay Acunzo · 2026-06-18 source ↗ Reader 🖥️
  AI slop, infinite online advice and content noise are all the same thing — secondhand information, knowing *about* — whereas the durable, non-commoditizable value of any creator's work comes only from firsthand lived experience synthesized through a distinctive personal perspective, so the response to the AI era is to draw on your own life rather than let the noise convince you your experience doesn't matter.
• Brain Food: Intelligence vs. Wisdom Shane Parrish (FS) · 2026-05-24 source ↗ Reader 🖥️
  What separates effective lives from merely productive ones is the constant subordination of capability to selection — choosing the right goals, removing what others tolerate, refusing the social tax of false 'niceness' — so that intelligence is harnessed by wisdom rather than wasted on impressive but misdirected striving.
• SWLW #701: Gazing at the misty AI Frontier, From Vibe Coding to Agentic Engineering, and more. Oren Ellenbogen · 2026-05-18 Reader 🖥️
  In the agentic era, software leaders win by raising their ceiling — investing in fundamentals, original thinking, and emotionally-attuned leadership — rather than letting AI flatten everyone to the same elevated floor.
• Why cultivating agency matters more than cultivating skills in the AI era | Max Schoening (Head of Product, Notion) Lenny Rachitsky, Max Schoening · 2026-05-09 Reader 🖥️
  In the AI era, the differentiating factor for product builders is not skill acquisition but agency — the willingness to treat the world as malleable and to make things — and the products that endure are those built around a tiny, exceptionally good core rather than an accumulation of features.
• Lenny Rachitsky's Post Lenny Rachitsky · 2026-04-24 source ↗ Reader 🖥️
  Anthropic's Claude Code team ships at compressed (week-or-day-scale) timelines by merging the PM/engineer/designer roles around product taste, deliberately building and shipping products before the underlying model is fully capable so the gap can be tested the instant a better model ships, building its own internal tools with Claude Code rather than buying SaaS, and hiring people who respond to escalating production chaos with enthusiasm rather than stress.
• How Anthropic's product team moves faster than anyone else | Cat Wu (Head of Product, Claude Code) Lenny's Podcast · 2026-04-23 Reader 🖥️
  In an era of rapidly improving AI models, the PM role must shift from multi-quarter planning and cross-functional coordination toward radical speed of execution, product taste, and the ability to continuously adapt the product harness to extract maximum value from each new model generation.
• Designing the AI-native engineering organization Engineering Enablement · 2026-04-20 Reader 🖥️
  Engineering organizations should not restructure their org charts in response to AI, but should fundamentally change how work happens within them — compressing team sizes, shortening planning horizons, shifting engineer time from create/operate to plan/validate, and redefining engineering skill around agency and product judgment rather than code production.
• The founder of Postman says you have to kill your existing org chart Ivan Burazin · 2026-04-10 Reader 🖥️
  Traditional organizational hierarchies must be dismantled in favor of flat, builder-centric structures where everyone — including PMs and designers — ships directly, and management exists to cultivate judgment rather than coordinate through layers.
• Research-Driven Agents: What Happens When Your Agent Reads Before It Codes Alex Kim · 2026-04-09 source ↗ Reader 🖥️
  Coding agents generate better optimizations when they read papers and study competing projects before touching code, because a codebase tells you what the code does but not why it's slow or what alternatives exist outside the repo — and the quality of an agent's hypotheses is bounded by the quality of its context.
• What happens when the coding becomes the least interesting part of the work Obie Fernandez · 2025-12-12 source ↗ Reader 🖥️
  Senior engineering value has always lived in mostly-tacit, pre-code judgment (pattern-recognition, abstraction level, blast radius, reversibility) rather than typing, which is why coding agents cannot yet substitute for it — but once senior engineers and their employers realize this, the organizational result will be a drastic contraction toward small teams of highly accountable, AI-empowered seniors who absorb roles that used to belong to juniors, PMs, business analysts, and designers.
• Eight years of wanting, three months of building with AI Lalit Maganti · 2025-03-01 source ↗ Reader 🖥️
  AI coding agents are a genuine force multiplier for implementation and last-mile polish, and can unlock projects that would otherwise die from the intersection of difficulty and tedium, but they are a dangerous substitute for design, architecture, and taste -- the developer must retain ownership of the soul of the software or risk building a codebase they cannot reason about.
• Agents don't know what good looks like. And that's exactly the problem. Luca Mezzalira · 2025-01-01 source ↗ Reader 🖥️
  Agentic AI systems lack the judgment to distinguish correct behavior from merely passing checks, which means that deterministic architectural guardrails — fitness functions, capability verification, explicit boundary definitions — are not optional but imperative; and the sociotechnical readiness of the humans and organizations that must own the generated systems is the overlooked bottleneck.
• 90% of My Skills Are Now Worth $0 Kent Beck · 2023-04-19 source ↗ Reader 🖥️
  A technological revolution has just made roughly 90% of Beck's professional skills economically worthless by making their output free, while multiplying the leverage of the remaining ~10% (knowing what to ask for, and judging the result) by a thousand, so the only sane response is to recalibrate which skills to invest in by rapidly trying many things rather than reasoning it out in advance.
• How to write with style Kurt Vonnegut · 1982 Reader 🖥️
  Good writing style is not verbal ornament but the honest self-revelation of a writer who genuinely cares about a worthwhile subject and respects the reader — and it is achieved not through games with language but through caring about content, sounding like the actual person you are, keeping language simple, and cutting anything that does not illuminate the subject.

Propositions touching this concept (65)

By publication month, oldest first (extraction month when no publication date is known).

1982

• implicitClarity is a moral obligation, not merely an aesthetic preference: to write carelessly or obscurely is to disrespect the reader, which makes legibility an ethical duty rather than a matter of taste.
  How to write with style · p9

2023-04

• explicitThe remaining skills that survive devaluation — chiefly the judgment to know what to ask the tool for and to recognize a good output — have their leverage multiplied by roughly three orders of magnitude.
  90% of My Skills Are Now Worth $0 · p2
• implicitA skill's resistance to AI devaluation correlates with how much of its value already came from taste/judgment versus from mechanical execution.
  90% of My Skills Are Now Worth $0 · p7

2024-10

• explicitAI adoption benefits organizational performance (+2.3% per 25% increase) and team performance (+1.4%) but shows no clear association with product performance, because team/org performance depend on communication, knowledge-sharing and decision-making bottlenecks that AI eases, while product success depends more on taste, empathy, and user-experience judgment that AI does not yet substitute for.
  2024_final_dora_report · p4
• explicitcorroboratedProgramming is already, and will keep becoming, more fun rather than less, as AI strips out boilerplate and 'cruft' and shifts the remaining human work toward taste, creative judgment and high-bandwidth intent communication (e.g. pressing Tab while 'injecting intent' continuously) rather than low-value carefulness and low-bandwidth text editing.
  Transcript for Cursor Team: Future of Programming with AI | Lex Fridman Podcast #447 · p31

2025-01

• explicit⚡ disputed & corroboratedCurrent agentic AI operates at the novice/advanced-beginner level on the Dreyfus scale: it can follow and adapt recipes but cannot exercise professional judgment about why a solution is appropriate.
  Agents don't know what good looks like. And that's exactly the problem. · p1

2025-03

• implicitcorroboratedThe ability to refactor cheaply with AI only has value if the developer has taste; without taste, cheap refactoring just rearranges confusion.
  Eight years of wanting, three months of building with AI · p11

2025-10

• explicitBig career outcomes have a luck component he is comfortable admitting — choosing Oracle over Google on the 2003 MIT Silicon Valley trip is his canonical self-deprecation about taste.
  1) Torbjørn Larsen: Det nærmeste vi kommer Steve Jobs · p28

2025-12

• explicitThe real initial effort in programming happens before the first line of code is written — pattern-recognition against years of experience finds the solution — making the actual typing/coding mostly follow-through, which is why coding agents are a game-changer for someone whose value was already in that pre-code judgment.
  What happens when the coding becomes the least interesting part of the work · p1
• explicitSenior engineering judgment runs on a largely tacit 'silent checklist' about shape, direction, and consequences (what kind of change is this, does it belong near the surface or deep in the system, what's the blast radius, is it reversible) rather than about code syntax — and teams rarely document or standardize this because it's hard to pin down and because talking about patterns/frameworks is easier than talking about instinct and timing.
  What happens when the coding becomes the least interesting part of the work · p2
• explicitWhether coding agents feel liberating or beside the point to a given engineer depends on which part of programming they actually valued: those who enjoy the physical act of typing code experience agents as irrelevant, while those whose value is judgment, tradeoffs, and intent experience agents as clearing space for exactly what they care about.
  What happens when the coding becomes the least interesting part of the work · p5
• implicitWhat people care about most in their job is also, in the author's framing, exactly the part most exposed to automation for those who care about typing rather than judgment — so the apparent safety of 'that's not what I care about' is actually a marker of vulnerability, not insulation.
  What happens when the coding becomes the least interesting part of the work · p7
• implicitcontestedThe argument assumes the organization's current supply of senior engineers — and the irreplaceable tacit judgment they individually hold — is a stable, already-adequate resource that can simply be redeployed more efficiently; it does not consider where the next generation of seniors will come from once the junior/mid roles that historically served as the training ground for that same tacit judgment are the first ones eliminated.
  What happens when the coding becomes the least interesting part of the work · p8

2026-02

• implicitcorroboratedImplicit: the 'AI replaces juniors' panic is misframed — the durable bottleneck is *evaluation capacity* (smak/judgment), which is hard to learn without doing the boring work that AI now hides; nobody on this panel has solved that.
  11) AI-spesial med Vilde Opsal og Anders Sveen · p23
• explicitHuman taste and judgment get captured once — through review comments, refactoring PRs, and user-facing bug reports — and are then enforced continuously and mechanically on every subsequent line of code, rather than being re-applied by a human reviewer each time a similar case recurs.
  Harness engineering: leveraging Codex in an agent-first world · p15

2026-04

• explicitcorroboratedManual QA by the developer is non-negotiable; it is where developer taste is imposed back onto the codebase and where new Kanban issues are generated; teams that automate QA produce 'slop'.
  [FULL WORKSHOP] AI Coding For Real Engineers - Matt Pocock, AI Hero · p15
• implicitestablishedWithout human eyes, AI-built products lack taste; taste cannot be automated.
  [FULL WORKSHOP] AI Coding For Real Engineers - Matt Pocock, AI Hero · p27
• explicit⚡ disputed & corroboratedThe future engineering skill profile prioritizes agency, product judgment, and generalism over deep technical specialization
  Designing the AI-native engineering organization · p8
• explicitcorroboratedAs code becomes much cheaper to write, deciding what to write (product taste) becomes the most valuable skill.
  How Anthropic's product team moves faster than anyone else | Cat Wu (Head of Product, Claude Code) · p3
• explicitManagement's purpose is no longer coordination or oversight but the development of better judgment.
  The founder of Postman says you have to kill your existing org chart · p5
• explicitOn the Claude Code team, almost all PMs have either been engineers or ship code themselves, and designers have been front-end engineers; many engineers go end-to-end from seeing user feedback to shipping a product within a week, without a PM involved -- the roles are merging and the most valuable skill is product taste, not job title.
  Lenny Rachitsky's Post · p3

2026-05

• explicitcorroboratedTaste is the ability to predict how an in-group will react to an idea, and it is built through iterative reps with feedback — not innate talent.
  Why cultivating agency matters more than cultivating skills in the AI era | Max Schoening (Head of Product, Notion) · p13
• explicitcontestedTaste may not be an enduring human advantage over AI — the input-idea-to-reaction loop resembles backpropagation.
  Why cultivating agency matters more than cultivating skills in the AI era | Max Schoening (Head of Product, Notion) · p14
• explicitdevelopedGut judgment that contradicts a rationalized head decision is a reliable signal that should usually be heeded
  SWLW #701: Gazing at the misty AI Frontier, From Vibe Coding to Agentic Engineering, and more. · p8

2026-06

• explicitdevelopedNadella defines token capital (a firm's owned AI capability — models, agents, traces, evals, workflow memory, internal loops) and human capital (its people's judgment, taste, relationships, context, pattern recognition) and argues the two should compound, so humans grow more valuable, not less, as machines do.
  What is the harder human-capital problem beneath token capital? · p1
• explicitAI does not create good engineering judgment or product instinct — it accelerates whatever habits already exist, amplifying strong taste and discipline but also amplifying weak habits (producing exactly what was asked without considering the outcome) louder, producing a distinct failure mode: output that is technically good but product-wrong.
  We don't write code anymore · p6
• explicitEven an algorithm that correctly models taste as changing over time still under-explores in practice, because its evaluation window (weeks to months) is far shorter than the multi-year timescale over which aesthetic tastes actually evolve.
  Recommendation algorithms might be making your entertainment boring, new research suggests · p5
• explicitcorroboratedThere are no red/green unit tests for taste: when there is no ground truth (as with 'which points of interest are interesting'), verification becomes hard to reason about.
  You can't unit test for taste · p1
• explicitThe same LLM nevertheless has useful 'taste' — subjective judgment latent in its weights — that lifts the more interesting items when objective signals (wiki language counts) over-reward noise like auto-translated small towns.
  You can't unit test for taste · p4
• implicitcorroboratedTaste is an irreducibly human faculty here: AI can supply raw material and even its own latent ratings, but a person must own the final judgment because no automated test can render it.
  You can't unit test for taste · p11
• implicitcorroboratedTaste is exercised largely as restraint and proportion — what to filter out, which feature codes to include, spreading points evenly rather than showing every town — so quality is as much about what you leave out as what you include.
  You can't unit test for taste · p13
• implicitTaste is developed by iteration against the messy particular: the scale of per-route variance 'didn't occur to him until we got this far,' not something specifiable up front.
  You can't unit test for taste · p14
• implicitcorroboratedThe untestable-taste problem looks confined to 'messy' data only because code has a ground truth; wherever AI generates output in domains without one — or where design quality itself has no red/green test — the same unverifiability applies. He stops short of turning this inference back onto AI-generated code.
  You can't unit test for taste · p15
• implicitcontestedTrusting the LLM's latent 'taste' is itself an unverifiable taste judgment: he can doubt a rating (Reykjavik = 10) but cannot check it against a ground truth, so delegating subjective scoring to a model is an act of faith that may launder its training-distribution bias — the very anglophone-Wikipedia bias he flagged a section earlier.
  You can't unit test for taste · p16

2026-07

• explicitWhat atrophies under AI is no longer manual hand-skill but thinking itself — writing, analysis, judgment — and unlike a wobbling plane, there is no alarm that warns your judgment has rusted.
  Det skal være vanskelig · p4
• explicitJudgment can only be built by doing the work: those who best evaluated AI's answers were those who had done the tasks manually for years, and greater trust in AI correlates with less critical thinking (Microsoft/Carnegie Mellon, 319 knowledge workers).
  Det skal være vanskelig · p11
• explicitAs AI absorbs more of the coding itself, distinctly human capacities — empathy, ethics, collaboration, judgment — increasingly become the main differentiator of an engineer's productivity, more than the code they personally write.
  Presentation: Getting Rid of LeetCode Interviews in the World of AI · p7
• explicitThe scarce resource in the agent era is judgment, not solving: when an agent can solve almost anything given an answer key, what becomes valuable is the judgment to decide which problems agents should solve and whether they actually got it right.
  Addy Osmani’s Post · p1
• explicitcorroboratedAn engineer's most durable edge is taste — earned pattern-matching that lets you tell good from great; the real risk of AI is not that it writes bad code but that engineers lose the ability to tell, so judgment must be continually renewed ('keep your alpha ahead of the models') through deliberate reps.
  Addy Osmani’s Post · p3
• implicitThe scheme presumes the marginal value of extra effort is discernible enough — by cultivated judgment, even without a measurable test — that a person can reliably locate 'good enough' and recognize the remainder as decoration.
  Why "good enough" is the whole point of 80/20 · p10
• implicitThe judgment of 'what doesn't earn its keep' can be soundly outsourced to an outside reader — a mentor or an AI — with that reader's taste taken as trustworthy; the risk that a borrowed (especially machine) read is itself unverifiable, bias-laden, or corrosive of one's own judgment is not considered.
  Why "good enough" is the whole point of 80/20 · p11
• explicitAutomating routine work (spotting weak arguments, interpreting ambiguous data, anticipating stakeholder concerns) removes the everyday repetition that used to build employees' judgment, pattern recognition, and intuition.
  Why Mentoring Matters More in the AI Era · p1
• explicitcontestedIf the on-ramp for professional judgment disappears, the leadership pipeline disappears with it, and on-demand learning platforms or one-off training sessions cannot transfer judgment, pattern recognition, or tacit knowledge.
  Why Mentoring Matters More in the AI Era · p2
• explicitEffective mentoring must be intentionally architected around explicitly named competencies (critical thinking, judgment, pattern recognition, proactive communication, stakeholder management, prioritization, navigating ambiguity), rather than left to develop informally.
  Why Mentoring Matters More in the AI Era · p4
• explicitSeniority does not automatically confer coaching or mentoring skill; mentors need training in question-led ('coaching') style versus directive style, and in creating psychological safety, so they teach judgment rather than just correct output.
  Why Mentoring Matters More in the AI Era · p7
• explicitSenior leaders can transmit otherwise-invisible judgment by narrating their decision-making in real time — before and after high-stakes moments — turning tacit expertise into an observable, teachable process.
  Why Mentoring Matters More in the AI Era · p9
• implicitcontestedThe judgment, pattern recognition, and tacit knowledge mentoring aims to build are capacities AI itself cannot substitute for or teach — they can only be transmitted human-to-human, through exposure, narration, and structured relationship.
  Why Mentoring Matters More in the AI Era · p12
• implicitThere is an effectively stable, undiminished reservoir of senior-leader judgment and expertise available to mentor from — the article assumes the supply being tapped is itself untouched by the same automation trend that is de-skilling juniors.
  Why Mentoring Matters More in the AI Era · p13
• explicitAs it becomes easier to build things, the scarce and important work shifts to figuring out what you should be building in the first place — so taste matters more, not less.
  Adam Mosseri: AI is a tailwind for authenticity · p1
• explicitcontestedTaste is 'much more difficult to imagine being automated away', so despite generalists, engineers and product staff now doing design work, Mosseri is 'long on designers' because they tend to have taste.
  Adam Mosseri: AI is a tailwind for authenticity · p2
• explicitcontestedHuman brains stay most valuable in taste, judgment, vision and strategy: you don't ask an AI to originate a strategy (only to give feedback within bounds you set), and directing AI 'looks more like management' — defining what success looks like, choosing how prescriptive to be, and giving feedback along the way.
  Adam Mosseri: AI is a tailwind for authenticity · p16
• implicitImplicit: the value of a person's work is migrating from producing output to exercising the human faculties AI cannot supply — taste, authenticity, judgment and the ability to steer — so both individual careers and team designs should be reorganized around those faculties.
  Adam Mosseri: AI is a tailwind for authenticity · p25
• explicitThe proven/better/new framework raises a product's odds of success: master what is already proven on the exact platform, audience, and experience (copying it legally and with taste), add a 'better' that 10 out of 10 existing users would endorse, and add exactly one 'new' idea that is expected to fail but gives people a reason to try.
  The hidden pattern behind successful products | Mark Pincus (Founder of Zynga) · p2
• explicitThe best product makers copy so tastefully that the derivation is invisible; consumer taste itself resists a naked copy, so the copied element must be combined with something genuinely better or new to earn a switch.
  The hidden pattern behind successful products | Mark Pincus (Founder of Zynga) · p6
• implicitTacit product judgment ('vampire blood') is transferred not through documents but by proximity — a 'teaching hospital' of people in the room plus a rotating tech-assistant apprenticeship — which scales the founder's leadership almost for free.
  The hidden pattern behind successful products | Mark Pincus (Founder of Zynga) · p26
• explicitcorroboratedThe scarce, valuable, hard-to-teach human skills are now judgment, taste, and troubleshooting, and there is no known way to teach them without first letting people do the grunt work agents now absorb.
  We've Stopped Arguing About Whether AI Changes Software · p9
• implicitdevelopedThe decisive work of engineering leadership under AI — judgment about what to build, sound operating principles, and engineering rigor — is unchanged in essence; AI shifts the magnitude, economics, and tooling, not the fundamentals.
  SWLW #710: Revised rules of Eng Leadership, When AI costs more than the engineer, and more. · p1
• explicitcorroboratedHuman judgment is more crucial now, not less — deciding what to build, which architecture, when to regenerate from scratch vs. reuse existing pieces, and how to manage tech debt — because you can do anything now but you can't do everything.
  SWLW #710: Revised rules of Eng Leadership, When AI costs more than the engineer, and more. · p12

2026-08

• explicitGood agent-driven work depends on two abilities: deep domain expertise (understanding the problem well enough to define a good outcome) and applied judgment (turning taste into a clear, testable plan).
  Agentic Skill Decay · p2

2026-09

• explicitHuman attributes like taste, accountability, judgment, clarity, and persistence are 'not available in abundance' and remain the differentiators AI cannot substitute.
  AI Is Changing the Rules of Entrepreneurship · p12
• explicitAnthropic recently removed roughly 80% of the Claude Code system prompt, including most tool-description examples, because examples had become mostly harmful — steering the model toward narrow imitation rather than judgment — as models got better at inferring intent.
  How Anthropic Builds And How Engineering Will Change Soon | Thariq Shihipar · p17
• explicitcontestedGetting tasteful/subjective output (e.g. visual design) depends on giving the model rich, structured references (an HTML or Figma file, not a screenshot) and on the human's own developed taste to judge when output is good enough — and that judgment skill becomes more, not less, valuable as models get more capable.
  How Anthropic Builds And How Engineering Will Change Soon | Thariq Shihipar · p19
• explicitConcrete, objective critic criteria (e.g. 'rank these 5 designs, 4 professional examples and 1 of ours, by polish and taste') produce more consistent evaluation than subjective criteria (e.g. 'judge if this looks beautiful, not AI-generated').
  How to turn your AI into a world-class designer · p7
• explicitTaste — the pre-metric judgment of what's coming, before there's evidence for it — is what identifies a coming alpha shift and senses a decay fading out; because any repeatable capability eventually decays into a commodity, taste is the durable competitive advantage, so it should be deliberately named, practiced in critique, and made explicit.
  Own the Outer Loop · p17
• implicitThe boundary between 'inside the system' (agent capability) and 'outside the system' (human agency) is treated as a stable, drawable line that doesn't itself need renegotiating as models get better — the essay never asks what happens when agent judgment starts encroaching on the 'outside' (decide/verify/approve) side.
  Own the Outer Loop · p24

Tensions

Open gaps (81)

• incompleteNo discussion of which quality signals resist deterministic proxying altogether. Mutation testing and unit tests can catch correctness regressions but not architectural coherence, elegance, or long-term comprehensibility — the very qualities ('clean', 'thoughtful') the post's own opening paragraph names as markers of good code.
  Addy Osmani's Post — read next: extract-garg-reducing-friction-ai-development.json (g2: Polanyi's paradox — the most important quality signals may be exactly the ones hardest to articulate as explicit constraints) and the software-design-fundamentals corpus (Ousterhout on deep modules, complexity) for what judgment stays irreducibly human.
• incompleteAgency is presented as the key differentiator but its uneven distribution is noted without exploring why. Structural barriers to agency (economic precarity, organizational culture, psychological safety, access to tools) are absent from the analysis. The advice to 'just make things' assumes access and psychological conditions that are themselves unevenly distributed.
  Why cultivating agency matters more than cultivating skills in the AI era | Max Schoening (Head of Product, Notion) — read next: Amartya Sen's 'Development as Freedom' on capability theory; research on psychological safety (Amy Edmondson) as a precondition for agency in organizations
• incompleteThe intelligence-saturation thesis (retina display analogy) is provocative but untested. Whether knowledge work has a clean saturation point or a long tail of edge cases requiring frontier intelligence is not examined. The analogy breaks if cognitive tasks are heterogeneous in their intelligence requirements.
  Why cultivating agency matters more than cultivating skills in the AI era | Max Schoening (Head of Product, Notion) — read next: Research on scaling laws and emergent capabilities in LLMs; Dario Amodei's 'Machines of Loving Grace' essay on where intelligence bottlenecks actually bind
• incompleteThe conversation repeatedly invokes 'quality' and 'craft' as things the industry is losing but offers no mechanism for measuring or incentivizing quality in an era where shipping speed is the dominant metric. The 'obviously good' standard is acknowledged to be subjective ('you know it when you see it').
  Why cultivating agency matters more than cultivating skills in the AI era | Max Schoening (Head of Product, Notion) — read next: John Ousterhout's 'A Philosophy of Software Design' on complexity as the enemy of quality; the Toyota Production System literature on building quality into the process rather than inspecting it afterward
• unasked questionThe essay uses the Dreyfus model to characterize agents as novice/advanced-beginner but never asks whether the Dreyfus model, developed for human skill acquisition, is even the right frame for AI systems that learn in fundamentally different ways. Agent capabilities may not map onto a linear human progression scale at all.
  Agents don't know what good looks like. And that's exactly the problem. — read next: Literature on AI capability evaluation frameworks that are not anthropomorphic (e.g., benchmarks, formal verification, capability elicitation research).
• implicit premiseThe argument that 'more context degrades output' treats current LLM context-handling limitations as a stable architectural constraint rather than a temporary engineering limitation. If context handling improves significantly, the 'deterministic guardrails' prescription may need rebalancing.
  Agents don't know what good looks like. And that's exactly the problem. — read next: Research on long-context model performance, retrieval-augmented generation, and structured context management.
• incompletePocock conflates two definitions of 'done': (a) feature shipped with QA-imposed taste, and (b) commit produced by the Ralph loop with passing tests. The transitions between (a) and (b) — especially the merge-and-conflict-resolution at the end of multiple parallel sandbox branches — are sketched (the Sandcastle merger agent) but not interrogated for what happens when the merger itself produces semantically broken code. This is exactly where small-PR discipline matters and where he admits he doesn't have the answer.
  [FULL WORKSHOP] AI Coding For Real Engineers - Matt Pocock, AI Hero — read next: Beck on incremental design and TCR; trunk-based development literature; the smallcode failure-mode catalog if it covers merge-stage failures.
• incompleteThe implicit premise that the developer can read agent output well enough to QA, code-review, and impose taste is enormous. Pocock states 'understanding code deeply is going to make you a much, much better developer and get more out of AI' and 'bad codebases make bad agents' — but does not address the developer who is themselves junior, distracted, or not the original architect. The workflow may be most valuable to seniors and least usable by the engineers who most need leverage.
  [FULL WORKSHOP] AI Coding For Real Engineers - Matt Pocock, AI Hero — read next: Marcus on code quality (in user's recent saves), Beck on apprenticeship/teaching contexts, Cutler on directly-responsible-individuals and skill distribution.
• uninformedNo mention of legal or IP questions raised by AI-generated business assets -- copyright status of AI-generated code, branding, and marketing copy, or liability when an autonomous agent like Felix independently spends capital and enters payment transactions.
  AI Is Changing the Rules of Entrepreneurship — read next: Legal scholarship on AI agent liability and the IP status of generative output; regulatory commentary on autonomous agents holding and spending capital.
• incompleteSycophancy and the 'illusion of progress' are named as core challenges, but the only prescribed mitigation is a general exhortation to 'leave their computers for the real world of commerce' -- no concrete practice (adversarial-critic prompting, real customer-validation gates, third-party audits) is offered, even though this is the crux of the 'expertise still matters' argument.
  AI Is Changing the Rules of Entrepreneurship — read next: Corpus material on adversarial-critic prompting patterns explicitly names the same unresolved failure mode -- see extract-eating-my-own-dog-food-ai-autonomy-framework#g5 ('model sycophancy despite hostile persona, critique that is adversarial in tone but generic in substance').
• incompleteAll three panelists represent large, well-resourced companies (Microsoft, 1Password, Atlassian). The advice assumes access to dedicated platform teams, AI champions, RL labs, and custom cost management tools. No consideration of how smaller organizations should approach these changes.
  Designing the AI-native engineering organization — read next: Seek perspectives from engineering leaders at startups or mid-size companies on AI adoption without dedicated enablement infrastructure
• illogicalThere is a tension between 'don't mandate AI' and 'the fastest learners will win.' If organizational learning speed is the competitive differentiator, laissez-faire adoption may be too slow. The panelists don't reconcile this.
  Designing the AI-native engineering organization — read next: Change management literature on balancing autonomy with urgency in technology adoption
• incompleteNo mechanism for how lived experience becomes transmissible work. The load-bearing premise — that experience filtered through a distinctive voice 'turns into meaning others can access', converting secondhand to firsthand — is asserted but never examined, and the recursion it creates (the reader receiving his transmitted meaning is then holding secondhand experience) is never addressed. This is the essay's central unexamined move and the point where an anti-hype reader will feel the truism.
  The best response to AI slop, infinite advice, and online noise is from Robin Williams — read next: Read the corpus's transmission-problem thread — ki-utvikling-og-folelser (how the next generation acquires craft), you-cant-unit-test-for-taste (where taste comes from / how it transmits) and svenson-attention-real-skill-ai-prompts (names the decisive skill, offers no method) — alongside Polanyi on tacit knowledge (does lived experience transmit at all?). Frame as infrastructure for getting lived experience *through* into distinctive, transmissible work, not as motivation.
• incompleteThe essay is secondhand advice, built on a secondhand fictional artifact, arguing against secondhand material; it survives only by silently reclassifying 'secondhand-through-a-distinctive-voice' as firsthand. That substitution relocates the real claim from 'firsthand experience matters' to 'distinctiveness of perspective matters' — two things that can come apart (and that AI-augmented work might also carry) — but the move is never made explicit or defended.
  The best response to AI slop, infinite advice, and online noise is from Robin Williams — read next: Read on performative / self-referential contradiction and test whether distinctiveness of voice, not novel firsthand experience, is the actual operative variable. Cross with eating-my-own-dog-food (human-originated seed vs prompt-generated framework) and franco-what-we-lose-when-ai-writes (voice as evidence of a Self) to see whether 'voice' can be specified independently of 'lived experience.'
• incompleteFirsthand vs secondhand is treated as a hard binary, but all human work recombines absorbed (secondhand) inputs — everyone's 'little life moments' are themselves saturated with others' content — and AI can also prompt, augment or occasion genuine firsthand experience. The essay admits no continuum and no role for AI as a spur to real experience rather than a substitute for it.
  The best response to AI slop, infinite advice, and online noise is from Robin Williams — read next: Read dawkins-claude-delusion (which flags exactly this sharp mimicry/genuine-experience binary as a weakness) and the taste article; probe where on the continuum 'your own LLM' actually sits, and where AI functions as a prompt for firsthand experience rather than a replacement of it.
• incomplete'It's all art' universalizes the firsthand-experience premium to every kind of knowledge work, colliding with the essay's own art-vs-science distinction (some work is a reproducible search for external truth — explicitly NOT art). The piece never marks where the claim stops applying, leaving an inspiring but unfalsifiable exhortation — a live risk given the creator/marketing register in which it is written.
  The best response to AI slop, infinite advice, and online noise is from Robin Williams — read next: Read Adler on separating a practical exhortation from a testable claim, and ask which categories of work the firsthand-experience premium genuinely governs versus where reproducible correctness — not distinctive voice — is what the audience needs (cf. the taste article's correctness-vs-quality axis, where 'correct' and 'good' come apart in the opposite direction).
• incompleteBeck appends a ChatGPT-generated pair of essays (pro and con) as 'bonus' evidence of ChatGPT's current writing quality, explicitly judging them as weak ('ChatGPT's writing sucks. No punchy analogies... No humor.') — yet he never reconciles this live counter-example with the post's own thesis that AI has driven his writing/wordsmithing skill to zero value. If the attached transcript is genuinely inferior Beck-style prose, that is evidence the 10%/90% split for a skill like 'writing with voice' may be misdrawn, or that the devaluation claim needs qualification by sub-skill (mechanical sentence-level editing vs. voice/style at the paragraph level) that the post doesn't draw.
  90% of My Skills Are Now Worth $0 — read next: Compare against later, more mature Beck writing on taste and judgment as the surviving human skill (e.g. the corpus's Taste & Judgment concept) to see whether Beck later resolves this tension by explicitly separating mechanical fluency from stylistic judgment.
• incompleteThe conversation is entirely from Anthropic's perspective. No competing viewpoints from OpenAI, Google, or independent practitioners are represented. The interview format means Lenny largely agrees with and amplifies Cat Wu's positions rather than stress-testing them.
  How Anthropic's product team moves faster than anyone else | Cat Wu (Head of Product, Claude Code) — read next: Read critical analyses of Anthropic's approach from outside observers; compare with product practices at other frontier labs.
• incompleteThe 'just do things' philosophy and extreme speed of shipping are presented as unqualified goods. The costs (product consistency, user confusion, bugs shipped) are acknowledged but treated as acceptable trade-offs without exploring when they become unacceptable — particularly for enterprise customers who need stability.
  How Anthropic's product team moves faster than anyone else | Cat Wu (Head of Product, Claude Code) — read next: Look at enterprise SaaS literature on release management and the cost of change fatigue for large organizations.
• uninformedThe discussion of Co-work replacing tedious PM tasks (decks, meeting prep, email triage) lacks any engagement with the quality ceiling problem: are these AI-generated outputs actually good enough for high-stakes contexts, or do they create a mediocrity trap?
  How Anthropic's product team moves faster than anyone else | Cat Wu (Head of Product, Claude Code) — read next: Seek empirical studies on AI-generated content quality in professional settings; read critiques of 'good enough' AI output.
• incompleteThe Open Claude decision is framed purely as a resource allocation choice. The community trust cost and the broader ecosystem implications of an AI company restricting how its model can be accessed through third-party tools are not explored.
  How Anthropic's product team moves faster than anyone else | Cat Wu (Head of Product, Claude Code) — read next: Read platform economics literature on ecosystem lock-in; review community responses and long-term effects of similar decisions by other platform companies.
• uninformedThe claim that human skills (empathy, ethics, collaboration, judgment) now matter more than code output is attributed secondhand to a single executive recruiter's remark in conversation, not examined against counter-evidence or defined more precisely — it is asserted as received wisdom rather than argued.
  Presentation: Getting Rid of LeetCode Interviews in the World of AI — read next: Primary research on which specific human capacities remain predictive of engineering performance as AI absorbs code production.
• unstated assumptionThe author assumes that the inability to externalize institutional knowledge and design context is a fundamental limitation of AI, not merely a tooling limitation. This conflates current AI capabilities with inherent constraints. Future models with persistent memory, codebase-scale context, and temporal reasoning could potentially bridge this gap.
  Eight years of wanting, three months of building with AI — read next: Read work on long-context models, AI memory architectures, and research on whether design taste can be formalized or learned (e.g., Christopher Alexander's pattern language as a possible bridge).
• omissionThe article does not discuss the economics of AI-assisted development beyond the $200/month Claude Max subscription. The 250 hours of personal time invested is mentioned but not analyzed in terms of opportunity cost or compared to hiring/contracting alternatives.
  Eight years of wanting, three months of building with AI — read next: Compare with analyses of AI coding tool ROI; consider the total cost including cognitive overhead and the discarded vibe-coding month.
• unexamined tensionThe author claims 'AI didn't just make the same project faster -- it changed what the project was' (adding editor extensions, docs, WASM playground). But he also claims the core value comes from the parser architecture, which AI was harmful for. This raises the question: did AI expand the project's scope in the right direction, or did the cheap last-mile features dilute focus on the hard core?
  Eight years of wanting, three months of building with AI — read next: Consider the 80/20 principle in open source: does breadth of features or depth of core reliability drive adoption?
• implicit premiseThe expertise-x-verifiability framework assumes that 'taste' and design judgment are inherently human and non-formalizable. This is stated as obvious but is actually a deep philosophical claim about the nature of aesthetic and design knowledge. If taste could be partially formalized (as design systems and API design guidelines attempt), the framework's boundaries would shift.
  Eight years of wanting, three months of building with AI — read next: Joshua Bloch's 'How to Design a Good API and Why It Matters'; API design guideline literature; whether LLMs can learn design heuristics from large corpora of well-designed APIs.
• incompleteThe whole argument rests on the meta-skills being 'ungradeable' and therefore unautomatable, but the piece asserts this durability without defending it — offering no account of why selecting, judging, and even 'answering for it' won't also be encroached on as models gain evaluative and agentic capability. 'Keep your alpha ahead of the models' concedes the edge is depreciating without saying how fast or how you would know when it is gone.
  Addy Osmani’s Post — read next: Arguments about which judgment tasks resist automation — verifiability, evals, the limits of LLM-as-judge — and the frontier of what agents can and cannot evaluate.
• incompleteThe piece predicts collapsing a 20-person team into 4-5 AI-empowered seniors who absorb the PM/BA/designer roles, but its own argument undercuts this: senior judgment is explicitly described as learned 'over the course of years and decades' by doing exactly the junior/mid work the piece says will disappear first ('even the most productive juniors and mids are going to start feeling like net negatives'). If that pipeline closes, the piece never explains where the next generation of seniors — the only people it trusts to hold this irreplaceable tacit knowledge — will come from.
  What happens when the coding becomes the least interesting part of the work — read next: extract-hbr-mentoring-matters-ai-era, which names this exact tension (the on-ramp for professional judgment disappearing alongside the flattening of the senior population expected to do the mentoring) as a structural risk to the leadership pipeline.
• uninformedThe claim that typing code 'feels like a waste of time' when billed at an expensive hourly rate treats the economics purely from the senior consultant's individual vantage; it doesn't address the same calculus for salaried engineers inside a company absorbing the AI spend, or whether clients will keep paying the same hourly rate once they also realize how much of the work is agent-executed.
  What happens when the coding becomes the least interesting part of the work
• incompleteThe boundary of 'essential human work' is asserted but under-drawn. Moreira says product people can safely leave AI to build software 'when safety and big concerns' are absent, while complex concurrent systems still need technical judgment — but she gives no criteria for classifying which systems warrant formal specification versus which can be left to unspecified AI generation. Without that decision procedure, 'defining correct behaviour is essential human work' cannot be operationalised: a team cannot tell where the line falls for their system.
  Podcast: Formal Methods for Every Engineer in an AI-Powered Future — read next: Read alongside risk-tiering / not-all-software-is-equal material — the autonomy-boundary and classification gap in extract-ai-funhouse-mirror-velocity-debt#g4, and the risk-tiering concept — which frame exactly the 'which changes get which rigor' judgment this talk leaves implicit.
• incompleteHeavy reliance on Khononov's externally maintained Modularity Skills pack as the thing that makes Sensor 3 work. The article does not examine: skill-pack versioning, drift, security/supply-chain assumptions, or what happens when the skill author changes their mind. The harness now has a third-party dependency at the level of taste, not just at the level of code.
  Three more static code analysis sensors — read next: Khononov's Learning Domain-Driven Design and his coupling talks; literature on prompt drift; the emerging 'skills-as-dependency' discussion (Osmani's agent-skills writing).
• incompleteSvenson prescribes outsourcing the 'is it done / what doesn't earn its keep' judgment to a mentor or an AI ('the fastest way to build confidence is to borrow someone else's'), but never addresses that an AI's 'taste' is unverifiable and re-encodes its training-distribution bias (the taste cluster's central empirical finding), nor the cognitive-surrender/dependency risk of borrowed confidence — the exact move the corpus flags as the point at which your own judgment stops forming and the outside read becomes your read. The essay recommends the very outsourcing its neighbours warn against.
  Why "good enough" is the whole point of 80/20 — read next: Read extract-you-cant-unit-test-for-taste (AI 'taste' is untrustworthy and launders anglophone-Wikipedia bias; the doubted Reykjavik=10 score) and the corpus's cognitive-surrender material on 'borrowed confidence'; ask when an outside read cultivates your own discernment versus when it substitutes for it and creates dependence on a non-stationary, biased arbiter.
• incompleteThe claim that 'the last 20% usually produces something no one will notice' is stated as a near-universal and paired with an outcome-only standard ('wins a case'), but the piece never marks the domains where the last, un-specifiable increment IS the value — craft, design, prose, brand, and high-consequence work (a safety-critical filing, an appellate argument) where refusing to stop is warranted rigor, not fear. Without a boundary, the advice over-generalizes from routine deliverables to work where 'good enough' forfeits the whole point.
  Why "good enough" is the whole point of 80/20 — read next: Read the taste cluster — extract-you-cant-unit-test-for-taste (correctness is not quality) and extract-how-to-write-with-style (Vonnegut's 'have the guts to cut' aims at what illuminates the subject, a taste target, not at good-enough) — plus Ousterhout on design quality; test whether 'good enough' is ever the wrong stopping rule and specify the tasks where it is.
• incompleteThe article calls for protected mentor time and more senior-leader involvement in daily narrated teaching, but doesn't engage the corpus's contemporaneous finding that organizations are simultaneously flattening management layers and pushing more hands-on technical work onto senior/advanced individuals — exactly the population this article expects to have spare capacity to mentor. The prescription and the structural trend may be in tension.
  Why Mentoring Matters More in the AI Era — read next: extract-postman-engineering-leadership-report-2026 (management-layer flattening + rising hands-on work among advanced/senior staff) and extract-you-dont-get-to-create-anything-beck-shoup#g4 (the apprenticeship channel under high manager-to-report ratios).
• uninformedThe prescriptions (real-time narration before/after meetings, 15-to-30-minute as-you-work reviews) implicitly assume synchronous, often co-located or at least highly available working relationships. There is no engagement with how this translates to distributed, asynchronous, or cross-timezone teams, where spontaneous narration is much harder to arrange.
  Why Mentoring Matters More in the AI Era — read next: Remote-work and asynchronous-collaboration literature; any corpus material on distributed team practices.
• incompleteThe Jobs analogy is named but not fully cashed out against its disanalogies. Jobs's reality distortion field was famously paired with cruelty, factional purges and product taste applied to consumer aesthetics. Larsen's framing is the opposite — toppdivisjon-ledelse runs on indre motivasjon, not fear; authenticity / easy-to-read-ness is the asset; the product is a public-sector welfare system. The conversation does not name these disanalogies, so the analogy carries more rhetorical work than it earns.
  1) Torbjørn Larsen: Det nærmeste vi kommer Steve Jobs — read next: Read Walter Isaacson's Jobs alongside Tracy Kidder's Soul of a New Machine and Marc Andreessen-tradition writing on charismatic founder leadership to surface the disanalogies; for the NAV case specifically, read Cutler on 'institutionalized overload' to see what indre-motivasjon-based mobilisation can hide. Connects to the gap in the Skjærpe ep 17 extract on transformations that fail in the liminal phase.
• incompleteThe people who can't pendle. Ulseth's account is sympathetic to both directions but assumes the listener has the option of swinging both ways. It does not address those who are only-leaders (lost technical currency) or only-deliverers (no taste for organisational complexity) — and whether they should be coached, paired, or left in their lane. There is a real possibility that pendling is a personality privilege as much as an organisational one.
  14) Øystein Ulseth pendler mellom å lede og levere — read next: Camille Fournier on technical-leadership transitions; Cutler on the costs of forcing technical contributors into managerial roles; Allen Holub on accountability and the limits of role-fluidity.
• uninformedNone of the conversation's strong quantitative-sounding claims (the autonomous-change percentage being unmeasured by design, the incident-rate increase, '100 times more testing code,' engineers 'not in the IDE for six months') are backed by cited internal eval or metrics data — they are informal, first-person estimates, despite Anthropic's own emphasis elsewhere in the corpus on evals as the standard of truth for claims about model/practice performance.
  How Anthropic Builds And How Engineering Will Change Soon | Thariq Shihipar — read next: Any published Anthropic internal engineering metrics or eval results on AI-assisted development adoption and outcomes, to check these estimates against measured data.
• uninformedEvery demonstrated result (seed-string variety, critic-loop convergence, the 10%-of-output-tokens cost claim, the calorie-tracker before/after) is a single-author, unreplicated anecdote with screenshots chosen by the author himself — there is no blind evaluation, no comparison group, and no measurement of how often these techniques fail or produce worse results than a baseline prompt.
  How to turn your AI into a world-class designer — read next: Look for independent practitioner writeups or controlled comparisons of prompt-injected-randomness and critic-subagent design techniques — cross-check against this corpus's multi-agent-orchestration cluster (e.g. extract-anthropics-three-agent-harness) for whether the generator/evaluator pattern holds up outside a single anecdote.
• incompleteThe article never reconciles its headline framing ('built in three prompts,' 'built in two prompts') with the seven distinct named techniques it goes on to describe — it's unclear how many techniques were actually stacked to produce the opening hero demos, or what the realistic cost/iteration/wall-clock budget of the full pipeline is for a team that isn't the author.
  How to turn your AI into a world-class designer — read next: Treat 'N prompts' claims skeptically until technique count and iteration count are reconciled; look for a follow-up or reader Q&A addressing total cost.
• incompleteThe essay is entirely of the print-essay era: it says nothing about digital, collaborative, or AI-assisted writing. Its whole model — a single author revealing a single self to a reader — does not touch what 'sounding like yourself' or 'style as self-revelation' mean when text can be machine-generated, co-written, or stylistically homogenized at scale.
  How to write with style — read next: Read against Franco 'What we lose when AI writes for us' (style/voice as the Self, and its convergence under AI), Acunzo 'The best response to AI slop' (distinctive firsthand voice as the non-commoditizable signal), and the taste article (correctness vs. quality when a machine produces the prose).
• incomplete'Sound like yourself' is under-specified to the point of circularity for a novice: it presupposes an already-articulate, trustworthy voice worth sounding like, and locates that voice in one's origin and childhood speech rather than in any developable skill. It offers no method for a beginner whose natural voice is inarticulate or undeveloped to find or grow one.
  How to write with style — read next: Read a pedagogy-of-voice source (Zinsser 'On Writing Well', Peter Elbow's 'Writing With Power') and the deliberate-practice literature on how voice is cultivated rather than inherited.
• incompleteThe simplicity argument generalizes from emotional profundity to all profundity. Every exemplar (Hamlet's soliloquy, Joyce's 'She was tired', Genesis) is emotional or narrative depth rendered plainly; none is a technically or conceptually irreducible subject. The claim that childlike simplicity fits 'the most profound subjects' may not survive contact with domains whose complexity is essential, not incidental.
  How to write with style — read next: Read Ousterhout on essential vs. accidental complexity and the difference between deep-simple and simplistic; explanation theory / technical-writing craft on rendering irreducible complexity without falsifying it.
• incompleteThe claim that non-standard voice is 'delightful' assumes a reader who is charmed by it ('like a very pretty girl with one eye green and one blue'), with no account of the many contexts where non-standard or regional voice is penalized rather than trusted. 'Sound like yourself' is treated as costless, which may reflect the privilege of a canonical author rather than the situation of a writer whose authentic voice carries real reputational cost.
  How to write with style — read next: Read sociolinguistics on prestige dialects and linguistic discrimination, and the literature on code-switching, to see where authentic voice is a liability rather than a delight.
• incompleteAuto research is demonstrated only on hyperparameter optimization — a narrow, well-defined search with clear metrics. No evidence or discussion of how it would work on research requiring genuine novelty, taste, or paradigm shifts.
  Skill Issue: Andrej Karpathy on Code Agents, AutoResearch, and the Loopy Era of AI — read next: Literature on AI-driven scientific discovery beyond optimization: can agents formulate novel hypotheses, not just test existing ones? See work on AI scientists (e.g., Sakana AI's 'The AI Scientist').
• unstated assumptionThe argument assumes that coordination, alignment, and context-sharing across teams are either unnecessary or automatically handled in flat structures. Large-scale organizations face coordination problems that are not solely caused by hierarchy but by the inherent complexity of distributed work.
  The founder of Postman says you have to kill your existing org chart — read next: Literature on coordination costs in flat organizations (e.g., Galbraith's information-processing theory of organization design); case studies of flat-org failures at scale.
• missing evidenceNo evidence is provided that PMs building APIs or designers shipping PRs produces better outcomes than the prior division of labor. The implicit claim is that directness equals quality, but this conflates shipping speed with product quality.
  The founder of Postman says you have to kill your existing org chart — read next: Research on specialization vs. generalization in knowledge work; outcomes data from companies that have adopted these structures.
• selection biasThe argument is attributed to the founder of Postman — a developer-tools company where engineering culture is dominant. The prescription may not transfer to organizations where the core work is not software (healthcare, logistics, manufacturing).
  The founder of Postman says you have to kill your existing org chart — read next: Compare with AI-era org design thinking in non-tech industries.
• unacknowledged tensionThe binary 'either you're building, or you're selling' erases important organizational functions: strategy, research, compliance, people development, customer success. These are neither building nor selling but may be essential.
  The founder of Postman says you have to kill your existing org chart — read next: Consider which organizational functions are genuinely automatable vs. which are being rhetorically dismissed.
• incompleteTaste and judgment are presented as durable human advantages 'much more difficult to imagine being automated away', but Mosseri gives no account of why they resist automation, and the opposing view — that taste is an in-group-reaction prediction loop AI can also learn to close — is never engaged.
  Adam Mosseri: AI is a tailwind for authenticity — read next: The argument (from the Schoening/Lenny conversation) that taste resembles backpropagation and may not be an enduring human advantage; theories of taste as learnable feedback vs. innate faculty.
• incompleteThe claim that agents 'can bring all the expertise' but 'can't bring judgement' is asserted without defining judgement or explaining why it is non-transferable, and it sits in unexamined tension with the author's own reliance on agents to extract engineering rigour.
  Notes from FOSE Europe — read next: Read on the distinction between taste/judgement and expertise in AI-augmented work, and on where human judgement remains load-bearing.
• illogicalThe newsletter's editorial-standards footer states it will not partner with products lacking 'evidence in human trials,' immediately after running a sponsored dog-food segment whose own cited evidence is food-composition/chemistry studies and a consumer taste test — not any controlled trial of health outcomes in dogs, and not human trials at all (the stated bar doesn't even apply to a pet product). The piece never notices or addresses this mismatch between its stated standard and its own content.
  override your stress response — read next: Look for veterinary-nutrition randomized trials comparing extrusion-processed vs. air-dried/gently-processed diets on actual long-term canine health outcomes, not just biomarker or composition studies, to test whether the underlying claim would survive the newsletter's own stated bar.
• incompleteThe essay proposes an 'accountability contract' (checklist understood, evidence, who's accountable, system status) but never specifies how any of these fields get populated from an actual hour-scale agentic run whose intermediate decisions are, by the essay's own admission, not fully recorded and cannot all be traced back to input tokens — leaving the central operational question (how do you produce the evidence you're claiming to own) unanswered.
  Own the Outer Loop — read next: Read Osmani's own 'Quality in the Harness' and 'Software Factories, Light and Dark' for the concrete evidence artifacts (tests, static analysis, review gates) he specifies elsewhere, and read agent-governance / decision-provenance literature for concrete audit-trail mechanisms.
• incompleteThe alpha/decay/taste argument asserts taste is the durable, non-automatable advantage without engaging the obvious rebuttal — visible elsewhere in this reader's own corpus (an AI judge applying a versioned rubric to score architecture judgment) — that judgment-before-evidence is itself an active automation target, not a stable human preserve; the essay's own test ('is it a capability? then it decays') is never applied to taste itself.
  Own the Outer Loop — read next: Read the corpus's agentic-fitness-functions entries on calibrated AI judges scoring judgment-heavy concerns, and reread Paul Graham's original 'taste' essay for whether he claims taste is permanently unautomatable or merely currently ahead of the frontier.
• incompleteParrish offers no operational test for distinguishing wisdom from intelligence. If wisdom is choosing the right goals, then choosing well is itself a kind of getting-what-you-want — meta-getting. The distinction risks collapsing into 'intelligence applied to first-order goals' vs. 'intelligence applied to goal-selection' — i.e., into a hierarchy of intelligence rather than a separate faculty. Without a test (e.g., regret-rate, time-horizon of evaluation, second-order satisfaction), the aphorism is a vibe, not a tool.
  Brain Food: Intelligence vs. Wisdom — read next: Annie Duke, *Thinking in Bets* and *Quit* (separating decision quality from outcome quality; resulting; expected value of changing course). Russ Roberts, *Wild Problems* (decisions whose value cannot be precomputed — the closest contemporary treatment of 'what's worth wanting'). Aristotle's *Nicomachean Ethics* on phronesis (practical wisdom) vs. sophia and techne. Harry Frankfurt on second-order desires ('Freedom of the Will and the Concept of a Person').
• incompleteSurvivor bias in the exemplars. Chung Ju-yung and Winston Weinberg are presented as evidence for diligence + speed + refusal-of-impossibility, but the population of people who pursued the same strategies and were destroyed by them is invisible in the newsletter. The implicit causal claim (mindset → outcome) is not separated from the role of context, capital, geopolitics, and luck.
  Brain Food: Intelligence vs. Wisdom — read next: Phil Rosenzweig, *The Halo Effect*. Michael Mauboussin, *The Success Equation* (luck vs. skill decomposition). Daniel Kahneman on regression to the mean and outcome bias.
• incompleteThe Nice/Kind reframing collapses an interpersonal question into an individualist one. 'Kindness is what you owe yourself' is rhetorically powerful but elides the relational and reputational costs of consistent time-defense, and the asymmetric power dynamics in which 'guarding your time' is available to some workers and not others.
  Brain Food: Intelligence vs. Wisdom — read next: Jenny Odell, *How to Do Nothing* (counter-perspective on attention and refusal as a class-coded privilege). Oliver Burkeman, *Four Thousand Weeks* (a less productivity-maximalist version of the same insight).
• incompleteThe newsletter equates 'wisdom' with selection of high-value goals but never engages the question of whose values define worth. The implicit value system (impact, achievement, output, building things that last) is presented as universal common sense rather than as a specific tradition. A reader who already shares the worldview will not notice; one who does not will find the wisdom/intelligence distinction culturally narrow.
  Brain Food: Intelligence vs. Wisdom — read next: Charles Taylor, *Sources of the Self*. Alasdair MacIntyre, *After Virtue* (on the disappearance of shared moral vocabulary that would make 'worth getting' intelligible). Agnes Callard on aspiration.
• illogicalThe 'failure is just feedback' frame depends on the video-game analogy, which is structurally disanalogous to most adult contexts: video games have low stakes, instant respawn, identical re-runs, and no observers whose evaluations persist. The aphorism is true under those conditions and progressively less true as stakes rise, irreversibility increases, and social audiences accumulate.
  Brain Food: Intelligence vs. Wisdom — read next: Nassim Taleb on ergodicity and the difference between ensemble averages and time averages — the conditions under which 'try again' is and is not available.
• incompleteThe piece offers no data on how typical Newton's experience is — he is an early-access, well-resourced tech journalist with a household connection to Anthropic, not a representative consumer; there is no discussion of skill floor, cost sensitivity, or failure/abandonment rate for a broader user base trying Glaze.
  Vibe coding has escaped the terminal — read next: Look for adoption/retention data on vibe-coding app-builders (Glaze, Wabi) beyond first-month enthusiast anecdotes — completion rates, subscription churn, or a more skeptical follow-up in the spirit of this corpus's 'AI Confidence Theater' theme, which treats vivid personal anecdote as a weak substitute for verifiable evidence.
• incompleteNewton flags the copyright/IP question about AI-generated Nightwing artwork but doesn't pursue what it implies for Glaze's business model, Raycast's legal exposure, or likely rightsholder response.
  Vibe coding has escaped the terminal — read next: Seek legal/industry analysis of IP liability for AI app-builders that call image-generation models, and how a marketplace like Glaze's store moderates for this.
• incompleteNo examination of the security/privacy architecture behind an app like Source Code, which merges phone contacts with LinkedIn data into a 'private, locally stored' CRM; Newton himself says he is 'still thinking through issues related to privacy and security,' but the piece never examines what guarantees Glaze actually provides around local storage, API-key handling, or data-exfiltration risk.
  Vibe coding has escaped the terminal — read next: Look for appsec-focused writing on the threat model of vibe-coded apps built by non-technical users that handle personal data (contacts, LinkedIn, API keys) — directly relevant to the reader's application-security interest.
• uninformedAs an outside-observer journalist rather than a practitioner, Newton doesn't engage with the mechanism underneath Glaze — what agent/model architecture it runs, how the 'circle an element to edit it' feature is implemented, or whether the generated code's quality or security is auditable. The piece is long on implications and 'vibes,' short on any rigorous look at what's actually happening under the hood.
  Vibe coding has escaped the terminal — read next: For the engineering-craft angle, look for a practitioner teardown of Glaze's (or Wabi's) underlying agent harness and tooling, analogous to a technical review of Claude Code internals.
• illogicalThe 'SaaS-pocalypse' and 'revenge' framing treats one person's delight in personalizing his own software as evidence for a market-level claim (a genuine threat to existing SaaS and App Store businesses), without addressing switching costs, the long-term maintenance burden of AI-generated one-off code, or whether most consumers actually want to build rather than buy.
  Vibe coding has escaped the terminal — read next: Look for a more rigorous build-vs-buy or market-structure analysis of personal software in the vibe-coding era, rather than a single enthusiast's month-one impression.
• uninformedThis is Lenny's secondhand, seven-bullet summary of a podcast interview whose full content is already extracted elsewhere in this corpus (extract-catherine-wu-claude-code-anthropic, 19+ propositions). Points 1, 2, 3, and 5 here restate claims already present in that primary extraction; points 4, 6, and 7 do not appear among that extraction's top propositions. It is unclear whether that reflects genuinely new material from the full conversation or simply a difference in what Lenny chose to highlight versus what the original extraction's author prioritized.
  Lenny Rachitsky's Post — read next: See extract-catherine-wu-claude-code-anthropic for the fuller primary-source extraction this post summarizes.
• incompletePoint 6 (abandoning Notion, Linear, and Figma in favor of custom Claude-Code-built tools plus Cowork and Slack) is presented as a pure productivity win, with no discussion of the switching costs, the maintenance burden of many bespoke internal tools built by non-engineers, or what is lost specifically in design handoff, accessibility review, or historical design-rationale tracking without a dedicated design tool like Figma.
  Lenny Rachitsky's Post
• incompleteThe article never states the model's functional forms, parameter values, or sensitivity analysis (how the appreciation curve is shaped, what noise levels were tested, how many Monte Carlo trials produced the reported effect sizes). Without these, a reader can't judge how robust the 'moderate noise helps' conclusion is to the specific modeling choices — a purely theoretical model's conclusions can be highly sensitive to functional-form assumptions.
  Recommendation algorithms might be making your entertainment boring, new research suggests — read next: Read the underlying paper directly — Samsun Knight, "Engagement-based curation and the evolution of taste," Journal of Cultural Economics (2026), https://doi.org/10.1007/s10824-026-09591-3 — for the actual model specification and robustness checks.
• incompleteThe article doesn't address how the agent decides whether it needs a research phase. The Liquid case didn't need one; the llama.cpp case did. But the agent couldn't have known this in advance. A decision heuristic ('when should I read before coding?') is the missing meta-skill that would make this approach self-directing rather than human-configured.
  Research-Driven Agents: What Happens When Your Agent Reads Before It Codes — read next: Metacognitive agent architectures; 'knowing what you don't know' in AI systems; confidence calibration research
• incompleteThe claim that well-optimized codebases (Redis, PostgreSQL, SQLite) are harder targets is mentioned in one sentence but not explored. This is actually the most interesting boundary: at what point has a codebase been optimized enough that agent-driven improvement becomes uneconomical? This would define the useful scope of the entire approach.
  Research-Driven Agents: What Happens When Your Agent Reads Before It Codes — read next: Performance engineering literature on diminishing returns; the 'last mile' problem in optimization
• incompleteThe research phase is described as reading papers and studying forks, but the article doesn't discuss how the agent evaluates which papers and forks are relevant. With thousands of arxiv papers and potentially dozens of forks, the agent needs a research strategy — and that strategy itself requires domain knowledge.
  Research-Driven Agents: What Happens When Your Agent Reads Before It Codes — read next: Information retrieval for technical literature; RAG architectures for agent research; the 'research taste' problem
• incompleteThe article is honest about agent overconfidence (the benchmark bug) but doesn't propose a solution. How should agents sanity-check their own instrumentation? This is a general agent reliability problem that extends far beyond optimization benchmarks.
  Research-Driven Agents: What Happens When Your Agent Reads Before It Codes — read next: Agent self-verification; Anthropic's work on AI oversight; the 'quis custodiet' problem in autonomous systems
• incompleteThe 'raise the floor but not the ceiling' claim is repeated approvingly but never operationalized. How does a leader, in practice, recognize ceiling-extending original work as opposed to confidently-presented model-mediated work? Without a heuristic, the framing becomes a flattering self-description rather than a usable filter.
  SWLW #701: Gazing at the misty AI Frontier, From Vibe Coding to Agentic Engineering, and more. — read next: Read Cedric Chin on tacit knowledge or Tyler Cowen on 'what counts as taste in an AI era' for concrete diagnostics; also Karpathy's full talk for his hiring criteria as a partial answer
• incompleteTwo of the curator's headline framings — 'AI raises the floor' and 'AI eats closed loops first' — are pulled toward each other but not reconciled. If AI raises the floor on closed-loop work, the implication is that closed-loop expertise becomes commodity labor while open-loop / relational / ethical work becomes the new premium. The newsletter implies this but does not state it, leaving leaders without an explicit instruction.
  SWLW #701: Gazing at the misty AI Frontier, From Vibe Coding to Agentic Engineering, and more. — read next: Elad Gil's full 'Misty AI Frontier' piece + a careful read of Yue Zhao's leadership argument against Christensen's disruption framing — the synthesis is more useful than either alone
• incompleteThe newsletter implicitly equates 'agentic engineering' (a craft skill) with 'agentic leadership' (a hiring/strategy task) but never separates them. Whether the same person should both code with agents and lead teams that code with agents — and whether the screening criteria are the same — is left ambiguous.
  SWLW #701: Gazing at the misty AI Frontier, From Vibe Coding to Agentic Engineering, and more. — read next: Bruno Fernandez-Ruiz's 'Engineering in the Compiler Age' slides (referenced here, worth a focused extraction) and Karpathy's actual hiring section to see how he separates IC craft from leadership selection
• incompleteThe 'Anthropic AI-resistant evaluation' story is treated as a positive exemplar, but the newsletter does not engage with the obvious tension: a long-horizon take-home where AI is permitted requires significant unpaid candidate time, which is a known equity problem. The curator endorses the design pattern without acknowledging its cost.
  SWLW #701: Gazing at the misty AI Frontier, From Vibe Coding to Agentic Engineering, and more. — read next: The original Anthropic write-up, plus any of the recent compensated-interview / pair-programming-screen literature, for a more honest cost-benefit picture
• illogicalThe endorsement of 'gut over head' (SahilBloom tweet) sits uneasily next to the endorsement of 'shared design concept before you code' (Pocock). The first valorizes unjustified judgment; the second valorizes explicit-and-shared rationalization before action. The newsletter treats both as obviously correct without noting that they pull in opposite directions for a leader deciding which to lean on in a given situation.
  SWLW #701: Gazing at the misty AI Frontier, From Vibe Coding to Agentic Engineering, and more. — read next: Daniel Kahneman on when intuition is trustworthy (closed environments with rapid feedback — i.e., precisely the closed loops AI is eating); Gary Klein's recognition-primed decision research for the conditions under which gut beats analysis
• incompleteThe issue diagnoses AI-era pressures (salary-scale economics, direction-less amplification, agent reliability, term-inflation) and poses them as open questions to the reader ('where does your company sit?', 'what would you do differently?') but never synthesizes them into an operating model — it stops short of what a leader should actually do.
  SWLW #710: Revised rules of Eng Leadership, When AI costs more than the engineer, and more. — read next: Will Larson's 'Revised Rules of Engineering Leadership' itself (the linked favorite), and John Cutler / engineering-strategy sources that convert principles into concrete decision rules.
• incompleteThe continuity framing (fundamentals and judgment are what matter, AI only raises their stakes) under-weights the possibility that some fundamentals themselves must change under AI — it treats 'fundamentals' as fixed rather than a live question.
  SWLW #710: Revised rules of Eng Leadership, When AI costs more than the engineer, and more. — read next: Sources debating whether the fundamentals taught today are still the right ones (e.g. Tobias Torrissen on juniors and fundamentals; 'The Future of Software Engineering' on mid-level retraining).
• incompleteHe asserts taste cannot be tested but never addresses how taste is developed or transmitted. He brings pre-existing taste to every judgment (the population-map problem, the anglophone-Wikipedia bias, the suspect Reykjavik=10) yet treats it as a given human endowment — silent on how a novice, a team, or the next generation would acquire or inherit it, especially once AI removes the on-ramp tasks that used to build it.
  You can't unit test for taste — read next: Read Polanyi on tacit knowledge and Dreyfus on skill acquisition (novice-to-expert); pair with the corpus's junior-apprenticeship thread (heis-09, heis-12, ki-utvikling-og-folelser) and Beck 'Tidy First?' / Ousterhout on design judgment as cultivated rather than innate. Ask: does taste transmit at all when AI removes the practice that used to form it?
• incompleteHe confines 'you can't unit test for taste' to messy data and treats code as the safely 'verifiable domain,' but never turns the lens onto AI-generated code itself, where tests pass yet design taste — simplicity, deep modules, restraint — has no red/green check. As agents generate more test-passing code, the untestable-taste problem migrates into the codebase, not away from it.
  You can't unit test for taste — read next: Read Ousterhout 'A Philosophy of Software Design,' Beck 'Tidy First?,' and Hickey 'Simple Made Easy' on design quality as unassertable; cross with corpus docs agents-dont-know-what-good-looks-like (behavioral vs capability verification) and cognitive-surrender (surface-correctness signals). Test the claim: is a green build ever evidence of good design?
• incompleteHe delegates the 'subjective' rating to the LLM's latent weights and finds it useful, but never examines whether that latent 'taste' is genuine judgment or averaged internet popularity — i.e., whether outsourcing scoring to a model launders the very anglophone-Wikipedia bias he diagnosed a section earlier into a 'subjective score' he can no longer see or audit.
  You can't unit test for taste — read next: Read on models as reflections of their training distribution (Bender et al.) and on popularity-as-proxy-for-quality critiques; probe the LLM ratings against non-anglophone sources to see whether the 'taste' is judgment or a re-encoding of who edits Wikipedia.