Perspectives With Josepha Haden Chomphosy  🔗 **Source:** [Perspectives With Josepha Haden Chomphosy](https://crossword.fm/s11/e10/) > [!note]+ Document Notes > tags: [wordpress, governance, ecosystem-thinking, open-source-leadership, ai-strategy, prioritization] > author: [Josepha Haden Chomphosy, with hosts Jonathan and Luke] > source: [Crossword (crossword.fm), Season 11 Episode 10] > reference: [https://crossword.fm/s11/e10/] > date created: 2026-08-07 00:00 > > > [!info] Generated analysis — not the author's words. Sections below are machine-written > > from the captured source. Only text inside `> [!quote]` blocks is verbatim from the author. > > Do not quote or attribute any other line to the named author. > > # 🏷️ Title: Perspectives with Josepha Haden Chomphosy (Crossword S11E10) > <!-- deep-read:sections start --> ## 🧭 Table of Contents - [🔗 References](#-references) - [📝 Summary](#-summary) - [🔍 Reader's Questions](#-readers-questions-courtney-2026-08-07) - [📖 Main Content](#-main-content) - [📚 References](#-references-primary) - [🗣️ Paraphrase](#️-paraphrase) - [📰 Newsletter](#-newsletter) - [✅ Action Items](#-action-items) - [🧪 Hypotheses](#-hypotheses) - [📊 Methodology](#-methodology) - [❓ Questions](#-questions) - [🔮 Future Research](#-future-research) - [💡 Implications](#-implications) - [🧠 Learning Styles Assessment](#-learning-styles-assessment) - [📊 Accessibility Assessment](#-accessibility-assessment) - [🌍 DEIB Assessment](#-deib-assessment) - [✅ Logical Validity](#-logical-validity) - [🛡️ DARVO Analysis](#️-darvo-analysis) - [🎯 Conclusion](#-conclusion) - [🧱 Refactor Appendix](#-refactor-appendix) - [🧰 Template Library](#-template-library) ### 🔗 References - Episode: https://crossword.fm/s11/e10/ (published 2026-08-05, ~7,912 words of transcript) - Josepha's site: https://josepha.blog/ - Book recommended in-episode: *Brave New Work*, Aaron Dignan - Mary Hubbard, "Big Picture Goals for 2026," Make/Project, 2026-01-23: https://make.wordpress.org/project/2026/01/23/big-picture-goals-for-2026/ - Make/Updates team reps (decision authority): https://make.wordpress.org/updates/team-reps/ - WP Tavern, Five for the Future "aspirational, not required" (2022-07-11): https://wptavern.com/wordpress-executive-director-says-5-in-five-for-the-future-program-is-aspirational-and-not-required ## 📝 Summary - **What happened.** Josepha Haden Chomphosy, WordPress project Executive Director 2019–2024, gives her first substantial post-departure interview on ecosystem thinking, drawing an extended analogy between running WordPress and serving on a city council. Hosts Jonathan and Luke use her framework to interrogate WordPress's current direction, particularly AI. - **Why it matters.** She supplies a compact, quotable prioritization test — WordPress users predictably want **secure, fast, easy to use** — and the hosts immediately apply it to current work and find it wanting. That test is the most portable artifact in the episode. - **Key stakeholders.** Core and Core AI teams; the Plugin Review team (Luke serves on it, and reports AI-driven submission volume); the Executive Director's office; hosting companies; the silent majority of site owners; agencies weighing WordPress against static/headless. - **Risks and tradeoffs.** The episode diagnoses a prioritization-legibility gap without naming any decision-maker, so nothing in it is actionable against a specific owner. Josepha explicitly recuses herself from evaluating current WordPress, which removes the one person best positioned to compare eras. - **Bottom line.** The strongest content is the three-predictables test and the "public toilets" framing for unglamorous infrastructure. The weakest is accountability: an episode substantially about who steers WordPress never names anyone who steers WordPress. ## 🔍 Reader's Questions (Courtney, 2026-08-07) > Four questions asked of this document specifically. Verification method is stated for each > claim, because several are claims about *absence*, which are cheap to test and easy to fake. > All word counts below were run with word-boundary regex against Reader's stored > `html_content` for this doc (id `01kzcwn7h2kx1snnwzxpzdaj62`), not against a re-scrape. ### Q1. The WordPress-specific bits, and WordPress problems in priority order **The WordPress content is roughly the middle third of the episode.** Everything else is municipal-government analogy. Extracted, the WordPress-specific claims are: Josepha's own test, stated as durable and predictable: > [!quote] I think that the three things that are always predictable for the future of WordPress are that people who are building websites or building apps on top of WordPress would like it to be more secure — both from a core perspective and everything that you can bolt onto core, like both of those need to be secure. I think they always want it to be fast — sometimes not faster, just not any slower is gonna count as well. And then also easy to use. Her definition of "easy" is unusually precise and worth keeping — it is a *journey* test, not a *control* test: > [!quote] is it easy for people to tell what comes in the basic bits? Is it easy to tell how far away their preferred future is? And is it easy for them to get there? Luke's rebuttal, which is the sharpest line in the episode because it negates all three at once: > [!quote] I'd suggest that these AI features that we're working on are making WordPress less secure, no faster, and harder to use. **Prioritized problem stack.** Ordered by *blast radius × how load-bearing the transcript evidence is*, and labelled by whether the transcript actually establishes the problem or merely gestures at it: | # | Problem | Priority | Evidence strength in transcript | |---|---|---|---| | 1 | **No legible prioritization mechanism.** Nobody outside can answer where WordPress is going, who decides, or how progress is measured. | Critical | **Strong** — Jonathan states it directly and at length | | 2 | **Attention allocates by interestingness, not by outcome.** "What's interesting to work on is kind of what gets the attention." | Critical | **Strong** — stated by Jonathan, independently corroborated by Josepha's "participating without an agenda, but also with an aim" | | 3 | **Unglamorous infrastructure is structurally underfunded** (the "public toilets" problem). | Critical | **Strong as a frame, weak as a list** — nobody names which WordPress systems are the toilets | | 4 | **Feedback is unrepresentative.** Loud contributors are mistaken for users; the passive majority is unreachable. | High | **Strong** — Josepha's most concrete, most self-implicating material | | 5 | **AI is arriving as an external event, not a governed choice.** Plugin Review absorbs the volume downstream. | High | **Medium** — asserted confidently; measured effects only for plugin submissions | | 6 | **Secure/fast/easy have no single owner** across core + plugins + themes + hosting, though users experience one product. | High | **Inferred** — Josepha names "core and everything you can bolt onto core"; the ownership gap is my inference, not her claim | | 7 | **Decentralization without shared institutions produces entropy.** | Medium-high | **Medium** — she says it plainly, cites the plugin directory as the counter-example | | 8 | **Responsive controls** — the only concrete feature gap named all episode (1 mention). | Medium | **Weak** — mentioned in passing, noted as already in progress | Two observations that matter more than the ranking: - **Security and performance get one mention each** (`security` ×1, `performance` ×1; `secure` ×6, `faster` ×4 — nearly all inside restatements of the same three-item list). For an episode whose thesis is that security and performance are the durable priorities, the episode itself allocates them almost no time. It enacts the problem it describes. - **The only occurrence of the word "governance" in the entire transcript refers to Josepha's municipal work**, not to WordPress. In an episode about who steers an ecosystem, WordPress governance is never named as such. ### Q2. Critiques of her time as leader? **Very few, and none load-bearing — but ChatGPT's "almost none" undercounts them.** She volunteers four self-critical items: 1. Municipal work gave her "a really wide lens that I don't think that I really had before" — a concession that her WordPress-era lens was narrower. 2. > [!quote] My very presence made people slightly less honest. This is the most genuinely self-implicating line in the episode: an admission that her leadership *method* had a structural defect she could not solve. 3. On being resisted by long-tenured contributors: > [!quote] and they're like, now listen, you. And I was like, yeah, I get it. I would hate me too. 4. On introducing centralized resources: "no one likes this answer, and WordPress didn't like it when I started doing it either." **But notice the shape.** All four are *"I was right and it was unpopular"* — self-criticism in form, self-validation in substance. Not one is "I was wrong." Verified by search: the words **"regret" (0)** and **"mistake" (0)** do not appear anywhere in the transcript. The missed follow-up is real and ChatGPT identifies it correctly. Luke asks the direct question — what were WordPress's public toilets? — and she answers: > [!quote] Gosh, what were the public toilets when I first brought that up? I can't remember what it was That is the natural door to *"so what did you prioritize, and did it work?"* Nobody walks through it. Also never raised, all verified absent by word-boundary search: **Five for the Future (0), Gutenberg (0), full-site editing (0), Automattic (0), WP Engine (0), trademark (0), fork (0), burnout (0), WordCamp (0)**. She never says the phrase "executive director" (0); the role surfaces once, obliquely, as "while I was the ED." ### Q3. Critiques of Mary and Matt? **Neither is named. Verified: "Matt" 0, "Mullenweg" 0, "Mary" 0, "Hubbard" 0.** (Naive substring counting returns false hits here — `matt` matches "matter," `mary` matches "summary." The zeroes above are word-boundary anchored.) **On Mary Hubbard.** The critique is unmistakably aimed at her portfolio while remaining formally deniable. Jonathan says he sees "no explicit evidence" that security, performance, and ease of use remain key priorities, and that AI is what's getting attention. Mary's published 2026 goals — **verified live at Make/Project, 2026-01-23** — list goal #2 as **"AI Everywhere, With Clear Guardrails and Benchmarks."** So the episode critiques the exact strategic posture of the sitting ED without ever putting her name, or her stated guardrails, on the table to be argued with. ChatGPT's characterization here is fair and its citation is accurate. **On Matt Mullenweg — ChatGPT calls this omission "more serious." I think that's the weakest part of its analysis, on two counts.** *First, a factual overstatement.* ChatGPT writes that project documentation "says Matt's approval is required for WordPress.org changes and that the executive director or Matt makes the final decision." I fetched the cited page. It actually says: > "the executive director (currently Mary Hubbard) or Matt will make the final decision about changes to the wordpress.org site." That is **disjunctive** — either party may decide. It is not a Matt veto. ChatGPT's first clause overstates its own source, and that clause is doing the work of establishing "exceptional formal control." The underlying point (final authority is concentrated and non-community) survives; the sharper version of it does not. *Second, and more important — the vault contradicts the framing.* **Crossword already ran a dedicated Matt Mullenweg interview** (covered in your vault at *3. Resources/Readwise/Articles/Five Takeaways from Matt Mullenweg's Crossword Interview.md*, The Repository, Rae Morey, ~2025-09-30). That episode pressed precisely the questions ChatGPT says this show avoids: the "WordPress.org belongs to me personally" remark, the bans, trademark defense, litigation and discovery costs, REST API in core, and his "judge me by outcomes" defense. A show that has already spent an episode putting those questions to Matt directly is not plausibly protecting him by leaving him out of a guest's ecosystem-theory episode. **Editorial sequencing explains the absence at least as well as evasion does** — and ChatGPT, working from the single transcript with no access to the show's back catalogue, could not have known that. ### Q4. Is there validity to the ChatGPT analysis? Is it "kiss-assing in reverse"? **Verdict: substantially valid on facts, over-confident on motive, and wrong on its single strongest charge.** What holds up — every absence claim I could test, plus both direct quotations: | ChatGPT claim | Status | |---|---| | Matt never named | ✅ verified, 0 mentions | | Mary never named | ✅ verified, 0 mentions | | No "regret" or "mistake" in transcript | ✅ verified, 0 each | | She can't recall the "public toilets," hosts move on | ✅ verified verbatim | | Responsive controls = only concrete deficiency named | ✅ verified, 1 mention | | Mary's 2026 goals include "AI Everywhere, With Clear Guardrails and Benchmarks" | ✅ verified live at source | | Five for the Future "aspirational, not required" quote | ✅ verified — Josepha, WP Tavern, 2022-07-11 | | Hosts challenge her ideas but not her record | ✅ accurate; Luke does push back hard on AI-as-requisite | What doesn't: - **"Matt's approval is required for .org changes"** — overstates the cited page (see Q3). - **"Roughly six years"** in leadership — she was ED **2019–2024**, about five. Minor, but it is the kind of inflation that makes a tenure argument sound heavier than it is. - **"Conspicuously absent... more serious"** on Matt — undercut by the show's own prior Matt episode. - **"Retrospective reputation laundering"** — this asserts *intent* from a *transcript*, which the transcript cannot support. **On "kiss-assing in reverse": no, and the phrase inverts what actually happened.** Reverse flattery would mean Josepha runs down her successors to make herself look better. She does the opposite — she recuses herself explicitly: > [!quote] I am not doing anything with it right now. So I actually don't know anything about the way that AI is in or not in WordPress at the moment. and volunteers unprompted: > [!quote] For what it's worth, number one thing from me is that I remain a strong supporter of WordPress. **The flattering-by-contrast effect is real, but it is produced by the hosts, not the guest.** The structure is: hosts diagnose present-day directionlessness → guest declines to comment on the present → guest supplies timeless principles the present appears to be violating. The net effect does burnish her, but the mechanism is her *discipline about scope* colliding with the hosts' agenda. Attributing that to laundering is reading a motive into a format. **The fairer criticism is one ChatGPT gets close to but doesn't land: this is an ideas interview being read as an accountability interview.** Crossword's "Perspectives" format books a guest for their framework, not their record. The legitimate complaint isn't that the hosts protected Josepha — it's that having a former ED on for the first substantial time post-departure was a scarce opportunity to compare eras, and the format spent it on municipal analogies. That's an opportunity cost, not a cover-up. **Net: ~75% of the ChatGPT analysis is verifiable and useful. Its problem-prioritization (Q1 half) is genuinely good and I largely agree with its ranking. Its accountability section is right about the facts and wrong about the motive, and its best-sounding line — the Matt omission — is the one your own vault refutes.** ## 📖 Main Content - **Ecosystem thinking vs systems thinking.** Systems thinking is legible ("if you've got a train, you need some rails"). Ecosystem thinking adds discoverability, extensibility, and the commercial substrate around a product — what she calls "the roads and bridges," and "the streets and future storefronts." - **The purpose of ecosystem work** is twofold: no user should need to know the right person to get an answer, and the maintainer should "make fertile ground for other, more brilliant people than you." She frames this as making "space for future genius." - **Municipal parallel.** She was appointed and then elected to a city council. Everything runs on 5-, 6-, or 10-year cycles — Luke immediately notes the resemblance to WordPress. The differentiator she highlights is *incentive structure*: open source incentives are collective and impact-rooted; municipal incentives are monetary across public, private, and philanthropic money. - **The fishbowl.** "When you're leading WordPress, you're the only WordPress in town" — no peer institution, total observation. In municipal work she is one of many towns, and must partner rather than be watched. - **The predictables doctrine.** Cities: amenities, better streets, better schools and parks. WordPress: secure, fast, easy to use. Her operating claim is that you can always know the top three things constituents will want regardless of what the future holds, so you build toward those. - **The maintenance-economics problem.** Streets in her 1939-built town were laid as "concrete on the dirt." Sometimes the fiscally correct move is to let a failing street fully fail before reconstructing. Residents hear "we're gonna let your stuff break" — and, she concedes, "technically that part is true." - **The communication gap.** "The distance between what you said, and what you thought you said, and what people actually took away with them is vast" — offered explicitly as something "we ran into a lot in WordPress." - **Fun work vs necessary work.** Her unpopular philosophical import to WordPress: "participating without an agenda, but also with an aim." Illustrated by a World Cup host city where everyone wants the tournament and someone has to ensure bathrooms and transit exist. - **AI exchange.** Luke frames AI enthusiasts as kids on e-bikes — "going way too fast... not aware, and completely inconsiderate." Josepha calls herself "pretty skeptical of AI" but says it's "essentially requisite at the moment." Luke rejects that on agency grounds. She lands on selective use: yes to diff-checking her own prolific writing and triaging whether a long read is worth reading; a hard no to AI in search, because "the majority of the world's internet experience is like a Facebook portal and an Instagram portal," because such systems are "very easy to trick," and because she needs statistics that are not "made up." - **Centralization.** Her most contrarian WordPress claim: getting further requires *some* centralized resources, "just to make it all make sense. And also for efficiency's sake, which is uncomfortable." Luke offers the plugin directory as the exemplar. Unbounded decentralization is "entropy — which can be a catastrophic state for a large organization." - **The passive majority.** 43% of the web means an enormous population passively experiencing WordPress. "Our next frontier was getting to those passive people, because we had made a tent for them and still couldn't get them." ### 🗒️ Notes - Format is a three-way conversation, not an interrogation; ~35 minutes in, Josepha jokes "You're asking the easy questions thirty minutes in." - Luke is on the Plugin Review team — the one participant with direct operational exposure to AI's effect on WordPress workflows. - Jonathan discloses he is migrating work off WordPress to static sites; this is n=1 but he is candid that it is his own behavior, not a measurement. - The episode ends on publishing friction: "you only see the final mile, which was already the hardest, most expensive mile." - Contact channels she names: Bluesky, Threads, LinkedIn. Explicitly not Twitter/X. josepha.blog currently hosts bread recipes. ## 📚 References (primary) - Crossword S11E10 transcript (primary): https://crossword.fm/s11/e10/ - Aaron Dignan, *Brave New Work* — her recommended entry point; her stated complexity threshold is "more than four stakeholders" - Make/Project 2026 goals (Mary Hubbard, 2026-01-23): https://make.wordpress.org/project/2026/01/23/big-picture-goals-for-2026/ - Make/Updates decision authority: https://make.wordpress.org/updates/team-reps/ - Vault cross-reference: `3. Resources/Readwise/Articles/Five Takeaways from Matt Mullenweg's Crossword Interview.md` (The Repository, Rae Morey) - Vault cross-reference: `1. Projects/llm-wiki-wp-governance/wiki/entities/Josepha Haden Chomphosy.md` (ED 2019–2024) ## 🗣️ Paraphrase - **Opening.** Asked to pitch ecosystem thinking at a dinner party, she distinguishes it from systems thinking: systems keep a complex organization running; ecosystems add the surrounding infrastructure that lets others discover, extend, and commercialize the product. - **On leaving WordPress.** She now sits on a city council, appointed then elected. The lesson: nothing meaningful in a city started two months ago or finishes in two months. - **On influence vs control.** You cannot force an ecosystem. You can know its shape, know your partners' ten-year horizons, and aim at the small set of outcomes constituents will predictably want. - **On WordPress's predictables.** Secure, fast, easy — where "easy" means a legible path from intent to done, and the doing may be delegated to a person, a vendor, or a model. - **On the present.** Jonathan cannot find evidence those remain the priorities; Luke argues AI work actively regresses all three; Josepha declines to assess current WordPress while affirming continued support for it. - **On maintenance.** The unglamorous work is what makes the fun work possible, and the person insisting on it will be disliked for it. - **On feedback.** As ED she spent much of her time asking for opinions and receiving feelings; her ask was always for the *because* — how many people, and what specifically would be better. - **On centralization.** Some shared, authoritative infrastructure is required for a decentralized ecosystem to function; the plugin directory is the proof. ## 📰 Newsletter **The three things everyone wants, and the awkward question of who's counting** Josepha Haden Chomphosy resurfaced on [Crossword](https://crossword.fm/s11/e10/) this week, and she brought a usefully blunt yardstick with her. Whatever else changes, she says, WordPress users predictably want three things: more secure, not slower, and easier to use. Bank that one — it fits on a sticky note and it survives roadmap churn. Then Luke Cavanagh-style bluntness arrives on schedule: the AI features currently absorbing the project's attention are, in his view, making WordPress "less secure, no faster, and harder to use." Three for three, against the very test the guest just supplied. Jonathan adds that he can't find explicit evidence those three remain priorities at all — and mentions, a little sheepishly, that he's been moving work to static sites. The best idea in the episode is the public toilets. Every ecosystem has infrastructure nobody wants to fund and everybody notices the instant it's missing. Josepha's line — that open source drifts toward the fun work — is the sort of thing that's obvious until you try to name WordPress's actual toilets. Asked to do exactly that, she can't remember what hers were. What's not in the episode is louder than what is. No Matt. No Mary. No Five for the Future, no Gutenberg, no regret, no mistake. An episode about who steers an ecosystem uses the word "governance" precisely once — about her city council. Worth your 45 minutes for the framework; bring your own accountability. **Links:** [the episode](https://crossword.fm/s11/e10/) · [Mary Hubbard's 2026 goals](https://make.wordpress.org/project/2026/01/23/big-picture-goals-for-2026/) · *Brave New Work* by Aaron Dignan ## ✅ Action Items - **DevRel** - `#action/h1` Turn "secure / not slower / easy" into a published, measurable release scorecard; ship it alongside each release post. - `#action/h2` Write the counter-piece: name WordPress's actual "public toilets" with owners and funding status. The episode poses the question and drops it. - `#action/h3` Office hours specifically for passive site owners, not contributors — the population Josepha names as the unreached frontier. - **OSPO** - `#action/h1` Policy & governance: publish an explicit map of which functions must be centralized (directory, security response, compatibility data, metrics) vs. genuinely decentralized. - `#action/h2` Security management: define cross-boundary ownership for "core + everything you can bolt onto core." - `#action/h3` Strategic alignment: tie AI investment to the three-predictables test as an explicit gate. - **Community** - `#action/h2` Build a feedback intake that demands the *because* — magnitude, affected population, proposed alternative — instead of sentiment. - `#action/h3` Structure listening sessions where participants talk to each other rather than to leadership, per her observation that presence suppresses honesty. - **WordPress (by team)** - Core: adopt an explicit no-regression bar for performance ("not any slower counts"). - Core Performance: publish trend dashboards spanning releases, not point benchmarks. - Core AI: apply the guardrails in the 2026 goals post as *gates*, publishing pass/fail per feature. - Core Program Team: publish decision records so direction is externally legible. - Mobile: verify AI surfaces don't regress mobile task completion. - Playground: build reproducible demos of the "how far is my preferred future" journey. - Community: meetup kits framed around the three predictables. - Training: Learn.WordPress path on evaluating AI features against user outcomes. - Marketing: stop leading with AI capability; lead with task completion. - Support: instrument which AI-adjacent questions generate the most tickets. - Design: close out responsive controls and publish the retrospective on why it took years. - Documentation: document where Core ends and the ecosystem begins — the boundary users cannot see. - Photos / TV: caption and transcribe governance content for the passive majority. - Polyglots: translate the prioritization scorecard, not just the marketing. - Accessibility: treat a11y as named infrastructure in the toilets inventory. - Meta: expose plugin-review queue depth publicly, given AI-driven submission volume. - CLI: ensure AI features have non-interactive equivalents. - Hosting: shared accountability for the performance floor. - Tide: revive automated compatibility signal as ecosystem-wide measurement. - Openverse: watch for AI-scraping pressure on the commons. - Themes / Plugins: publish AI-generated submission rates and rejection reasons. - **The WPOCC** - `#action/h2` This episode is a clean artifact for the collective's governance-legibility argument: a former ED describes the prioritization gap without naming a decision-maker. Log it. - **FAIR** - `#action/h2` Her centralization concession — that some authoritative shared infrastructure is required — is directly relevant to FAIR's federated-vs-central design debate. Cite it. ## 🧪 Hypotheses 1. If WordPress published a per-release scorecard against secure / not-slower / easy-to-use, external perception of "no clear direction" would measurably decline within three release cycles. *Testable via repeated ecosystem sentiment survey.* 2. AI-assisted plugin submissions have materially increased Plugin Review queue depth and time-to-decision since 2025. *Testable directly from Meta queue telemetry; Luke's team holds the data.* 3. Contributors self-select toward novel work over maintenance at a rate measurable in Trac/GitHub component activity, independent of stated priorities. *Testable by mapping commit volume to component "novelty" classification.* 4. The passive majority of site owners rank ease-of-use above AI features by a wide margin. *Testable by surveying via update-screen intercept rather than Make blogs.* 5. Interviews with former project leaders in "ideas" formats produce systematically fewer accountability questions than in "news" formats. *Testable by coding question types across a corpus of WordPress leadership interviews — including Crossword's own Matt episode as the contrast case.* 6. Ecosystems with explicit published decision rights show lower contributor churn than those without. *Cross-project comparison: WordPress vs. Drupal, Python, Rust governance models.* ## 📊 Methodology **Of the source:** none. This is an unstructured three-person conversational interview with no data collection, no sampling frame, and no measurement. Every empirical claim in it is anecdote or recollection. The one statistic cited — 43% of the web — is stated from memory and unsourced in-episode. Treat the entire transcript as testimony, not evidence. **Of this analysis:** claims about the transcript were verified by word-boundary regex (`\b`) over Reader's stored `html_content` for doc `01kzcwn7h2kx1snnwzxpzdaj62`, after `html.unescape()` and tag-stripping — deliberately not against a fresh scrape, so the counts describe the exact artifact in the library. Naive substring matching was rejected after it produced false positives (`matt`→"matter", `mary`→"summary"). External citations in the compared ChatGPT analysis were each fetched and read at source. Vault cross-references were located by filename and full-text grep across `~/Documents/2nd Brain`, excluding `Full Document Contents/`. ## ❓ Questions - What *were* WordPress's public toilets during 2019–2024, and which were funded? The episode asks and abandons this. - Did Josepha have decision authority, or accountability without authority? Given that Make/Updates documents final authority resting with the ED **or** Matt, the practical division during her tenure remains unexamined publicly. - What is the actual measured effect of AI submissions on Plugin Review throughput? Luke is sitting on the answer. - Who owns "secure, fast, easy" across the core/plugin/theme/host boundary that users experience as one product? - If the three predictables are genuinely predictable, why does no release communication report against them? - Is "ecosystem thinking" falsifiable? What observation would show it had failed? - Was the 2019–2024 era measurably better on these three axes? Nobody — including the person best placed to answer — is being asked. ## 🔮 Future Research - Longitudinal measurement of WordPress on the three predictables across 6.0→7.x, which would convert this episode's rhetoric into a testable record. - Comparative study of open source projects that publish decision rights vs. those that don't, keyed to contributor retention. - Study of the "passive majority" problem: what intervention actually reaches non-participating site owners at 43%-of-web scale? - Media analysis of post-departure leadership interviews across open source — do ideas formats systematically launder records? - Empirical work on whether AI tooling in CMSs improves or degrades task completion for non-expert users, which is precisely the claim Luke asserts and cannot support. ## 💡 Implications - **Practice.** The three-predictables test is immediately usable as a feature gate. Any team can ask of any proposal: does it make this more secure, not slower, and easier to complete a real task? - **Policy.** The centralization concession from a former ED is politically significant — it licenses shared authoritative infrastructure without conceding the ideological argument to central control. - **Security.** "Core and everything you can bolt onto core" is the right scope statement and the wrong ownership model; nobody owns the union. - **Education.** Her definition of "easy" — know what's included, know the gap, know how to close it — is a better curriculum design brief than most learning-path documents. - **Funding.** The public-toilets frame is the strongest available argument for funding maintenance, and it's wasted unless someone produces the actual inventory. ## 🧠 Learning Styles Assessment - **Visual.** Poorly served. Audio-first with no diagrams; the ecosystem/city analogy begs for a visual that doesn't exist. - **Aural.** Strongly served. Conversational, three distinct voices, natural turn-taking; the primary intended mode. - **Verbal.** Strongly served. Full transcript available; the metaphors (roads, toilets, e-bikes, potholes) are linguistically vivid and memorable. - **Kinesthetic.** Weak. Nothing to do or build; no exercises. - **Logical.** Moderate. There is a real argument structure (predictables → measurement → attention allocation), but it's never made explicit or diagrammed. - **Social.** Strongly served. The disagreement between Luke and Josepha over AI-as-requisite models productive dissent well. - **Solitary.** Well served via transcript for self-paced reading. - **Naturalistic.** Unusually well served — the entire framework is literally ecological, and the city/street/infrastructure analogies are concrete and environmental. - **Reading level.** Conversational, roughly grade 9–11. Sentences are long and clause-heavy (spoken register), but vocabulary is plain. Jargon requiring outside context: "ecosystem thinking," "Five for the Future" (never actually said), "plugin review," "ED." ## 📊 Accessibility Assessment Mapped to WCAG 2.2 AA, evaluating the published episode page. 1. **Perceivable — text alternatives.** ✅ A full transcript is published alongside audio, which is the single most important accessibility affordance for a podcast and is present here. 2. **Perceivable — time-based media.** ⚠️ Transcript is provided; no indication of synchronized captions or chapter markers. Reader's import also flagged that some embedded player content could not be imported, meaning the audio itself is behind an embed. 3. **Operable — keyboard.** ⚠️ Not verifiable from the captured artifact; depends on the Castos embedded player's controls. 4. **Operable — enough time.** ✅ Self-paced text; no timeouts. 5. **Understandable — readable.** ✅ Plain conversational language; speaker names are bolded and consistent, so the transcript is navigable by screen reader. 6. **Understandable — predictable.** ✅ Consistent speaker-label structure throughout. 7. **Robust — compatible.** ⚠️ Transcript uses semantic paragraphs with bolded speaker labels rather than a structured dialogue element; adequate but not ideal. No heading structure within the transcript body, so a screen-reader user cannot jump between topics across ~7,900 words. **Priority fix:** add headings or timestamps within the transcript. A 7,900-word unbroken transcript is technically accessible and practically unnavigable. ## 🌍 DEIB Assessment - **Cultural breadth.** Mixed. The municipal analogy is deeply US-specific — city councils, World Cup host cities, US housing inventory, US child-mortality statistics on e-bikes. Luke (Australian) has to interrupt to ask what "municipal government" means, which the episode handles gracefully and which usefully exposes the assumption. Josepha does check herself: "I don't know if you knew that, Luke — there's no reason for you to know." - **Inclusive language.** Good. No exclusionary language observed. She actively guards against ableist framing on e-bikes, insisting on assistive-technology access: "I am absolutely a hundred percent in favor of assistive technologies that help people have equal experiences," and warning against taking devices from "the people who need them or use them as primary transportation." - **Neurodiversity support.** Moderate. Full transcript supports non-auditory processing and re-reading. Against that, the conversation digresses heavily and the through-line is hard to hold; no summary, chapters, or takeaways are offered. - **Representation.** Limited. Three participants, all long-tenured in the same Western open source milieu. No one from the passive majority they spend the last third discussing. Notably, an episode about who WordPress fails to hear from contains no one who is hard to hear from. - **Psychological safety.** Strong, and the best DEIB feature of the episode. Luke asks permission to disagree ("Can I disagree with you?"), Josepha grants it, and the disagreement stays substantive and warm. Her framing of contribution capacity — echoing her documented position that Five for the Future's 5% is aspirational because "contribution to open source is a question and indication of privilege" — is equity-aware, though that argument is not made in this episode. ## ✅ Logical Validity **Strong reasoning** - The predictables argument is genuinely well-constructed: because you cannot predict the future, optimize for the small set of preferences that are invariant across futures. That's a sound decision rule under uncertainty and it transfers cleanly between cities and software. - The maintenance-economics passage is honest about its own unpalatability. She states the counterintuitive conclusion (let it fully fail), states the human cost, and concedes the residents' objection is literally true. That is intellectually fair. - The centralization argument is self-aware about its ideological cost — she flags it as uncomfortable rather than pretending it isn't. - The feedback critique is specific and actionable: not "people complain," but "tell me how many everyones, and tell me what the better button might be." **Fallacies** 1. **Appeal to inevitability / bandwagon** — Josepha: AI "is a tool that is essentially requisite at the moment. Like, it's required for people to use it." *Located:* AI exchange. *Why it fails:* widespread adoption doesn't establish necessity, and "required" is never grounded in any specific role or task. *Fix:* Luke supplies it in real time — he invokes agency and asks for the mechanism. She then retreats to selective use, which is the defensible position she should have started from. 2. **Weak analogy** — the AI ↔ kids-on-e-bikes comparison. *Located:* Luke's connective-tissue riff. *Why it fails:* the shared property is "moving fast and inconsiderate," which is an affect, not a mechanism; e-bike harm is physical, immediate, and measurable, while the alleged AI harm is diffuse and contested. Luke half-concedes: "I appreciate you both helping me to stretch this metaphor far beyond where it was meant to go." *Fix:* name the specific mechanism of harm (supply-chain risk in generated plugins, review-queue load) instead of the vibe. 3. **Unfalsifiable appeal to intuition** — Luke: people slower on AI "will turn out to be more employable, more useful, more valuable." *Located:* immediately after; he flags it himself — "it's a gut feeling, I can't actually give you facts and figures as to why." *Fix:* he is admirably transparent, but the claim should not then be used as an argument, which it is. 4. **Hasty generalization** — Luke: AI features are "making WordPress less secure, no faster, and harder to use." *Located:* the AI turn. *Why it fails:* three strong empirical claims, zero features named, zero measurements. It's the most quotable line in the episode and the least supported. *Fix:* name the feature and the regression. 5. **Anecdotal evidence** — Jonathan's migration to static sites as ecosystem signal. *Located:* his answer on whether WordPress is improving. *Mitigation:* he is honest that it's personal, but it still functions rhetorically as evidence of decline. 6. **False consensus** — "potholes are not partisan." *Located:* incentives discussion. *Why it fails:* the *desire* for good roads is non-partisan; *which* roads get funded first, and at whose expense, is exactly what is contested. The maxim smuggles agreement about allocation out of agreement about goals. 7. **Motte-and-bailey risk (structural, not a discrete instance)** — "ecosystem thinking" oscillates between an uncontroversial motte ("complex organizations need infrastructure") and an ambitious bailey ("this framework should govern WordPress prioritization"). No falsification condition is ever offered. 8. **Argument from silence, inverted (applies to the ChatGPT reading, not the episode)** — inferring reputation-laundering *intent* from what a transcript omits. Absence of accountability questions is evidence about the format; it is not evidence about motive. **Tone of voice.** Warm, self-deprecating, digressive, conflict-avoidant. Josepha is measured and generous, deploys humor to defuse ("I would hate me too," "my elevator pitch did not work"), and is scrupulous about staying inside her knowledge. The hosts are collegial and treat her as a peer, not a subject. The register is friendly-professional throughout, with genuine but affectionate disagreement. Notably: no defensiveness, no score-settling, and no criticism of any named person by anyone. ## 🛡️ DARVO Analysis **Overall: DARVO is largely absent, and it would be an overreach to claim otherwise.** No one is accused of anything, so there is no offense to deny, no accuser to attack, and no victim role to invert. Recording the components honestly: - **Deny.** ⚪ Not present. She denies no allegation because none is made. The closest structural analogue is *omission* rather than denial — the era-comparison question is never posed, so nothing is denied. Her "I actually don't know anything about the way that AI is in or not in WordPress" is a genuine scope limitation, not a deflection: she volunteers it before being pressed. - **Attack.** ⚪ Not present. Notably absent given the opportunity. She never criticizes successors, never names anyone, and preemptively affirms support for the project. Even the long-tenured contributors who resisted her are described with sympathy ("I would hate me too"), not as antagonists. - **Reverse victim and offender.** 🟡 **Faint, and worth naming precisely because it's the only component with any signal.** The retrospective framing consistently positions her as the person who saw correctly and was resisted: contributors preferred fun work, people gave feelings instead of evidence, long-tenured members pushed back, decentralization produced entropy. Each may be true. Cumulatively, they position leadership as acted-upon rather than accountable. This is a *mild* pattern and it falls well short of DARVO proper: there is no offense being reassigned, no accuser being recast, and she does concede one genuine structural failing of her own ("my very presence made people slightly less honest"). Best described as **self-serving retrospective framing**, not DARVO. Do not tag it `#darvo/reverse` in the graph; it does not meet the bar. ## 🎯 Conclusion **Bottom line.** The episode's durable contribution is a three-item prioritization test — secure, not slower, easy to complete a real task — plus the "public toilets" frame for unglamorous infrastructure. Its structural failure is that it diagnoses a prioritization and direction problem for an ecosystem while naming nobody who sets priorities or direction, and while its central witness explicitly recuses herself from the present. The ChatGPT critique you shared is largely right about what the transcript lacks and reasonably right about the WordPress problem ranking, but it overstates one of its sources on Matt's decision authority, inflates her tenure by a year, and — most importantly — treats the absence of Matt as evasion when this same show already devoted an entire episode to putting exactly those questions to him. "Kiss-assing in reverse" is the wrong diagnosis: the flattery-by-contrast is real, but it is manufactured by the hosts' framing colliding with the guest's genuine scope discipline, not by the guest running down her successors. **Next steps.** 1. Write the public-toilets inventory the episode refuses to write — WordPress's actual underfunded infrastructure, with owners. 2. Pull Plugin Review queue telemetry to test Luke's AI-volume claim empirically; it's the one assertion here with data behind it. 3. Cross-link into `llm-wiki-wp-governance` against `[[Josepha Haden Chomphosy]]`, `[[Mary Hubbard]]`, and the prior Crossword/Mullenweg source note. 4. Consider whether the three-predictables test belongs in The WPOCC's governance-legibility argument as a citable external yardstick. ## 🧱 Refactor Appendix ### Literature Split Pad #### L1 — The three predictables as a prioritization invariant `#atom/literature/h1` > [!quote] the three things that are always predictable for the future of WordPress are that people who are building websites or building apps on top of WordPress would like it to be more secure... I think they always want it to be fast — sometimes not faster, just not any slower is gonna count as well. And then also easy to use. Claim: under uncertainty, optimize for invariant preferences. Source: Josepha, Crossword S11E10. #### L2 — "Easy" is a journey property, not a control property `#atom/literature/h1` > [!quote] is it easy for people to tell what comes in the basic bits? Is it easy to tell how far away their preferred future is? And is it easy for them to get there? Claim: usability is legibility of the path from intent to outcome; execution may be delegated. #### L3 — Attention allocates by interestingness, not outcome `#atom/literature/h2` > [!quote] What's interesting to work on is kind of what gets the attention, and we're not necessarily connecting that to how it's impacting the ecosystem. Claim (Jonathan): open source has an attention-allocation failure decoupled from impact measurement. #### L4 — Leadership presence suppresses honest feedback `#atom/literature/h2` > [!quote] My very presence made people slightly less honest. Claim: the feedback channel is distorted by the authority of its collector — an argument for peer-to-peer listening formats. #### L5 — Decentralization without shared institutions is entropy `#atom/literature/h2` > [!quote] if you want to get much further than you were when I got there, you have to have some centralized resources, just to make it all make sense. And also for efficiency's sake, which is uncomfortable. Claim: some authoritative shared infrastructure is a precondition for a functioning decentralized ecosystem. Exemplar offered: the plugin directory. #### L6 — The passive majority is the unreached frontier `#atom/literature/h3` > [!quote] our next frontier was getting to those passive people, because we had made a tent for them and still couldn't get them. Claim: openness of participation does not produce representativeness of participation. #### L7 — AI features asserted to regress all three predictables `#atom/literature/h3` `#fallacy/hasty-generalization` > [!quote] I'd suggest that these AI features that we're working on are making WordPress less secure, no faster, and harder to use. Claim (Luke): unevidenced but falsifiable, and the most testable proposition in the episode. ### Atomic Split Pad - **Z1 — Ecosystem thinking `#atom/zettel/z1`.** Systems thinking plus the discoverability, extensibility, and commercial substrate around a product. "The roads and bridges." Threshold of applicability: more than four stakeholder types. - **Z2 — The public toilets problem `#atom/zettel/z1`.** Necessary infrastructure that nobody wants to fund and everybody notices only in its absence. Generalizes past open source to any commons. - **Z3 — Predictable-preference optimization `#atom/zettel/z1`.** When the future is unpredictable, build toward the preferences that hold across all futures. - **Z4 — Presence-distorted feedback `#atom/zettel/z1`.** The authority collecting feedback changes the feedback. Corollary: convene people to talk to each other, not to you. - **Z5 — Productive failure economics `#atom/zettel/z1`.** Sometimes the cheapest path is to let a failing asset fully fail before reconstruction — and the correct decision is the one that sounds most callous. - **Z6 — Participating without an agenda but with an aim `#atom/zettel/z1`.** Contribution should be outcome-directed without being self-interested. Her stated first philosophical import to WordPress; noted as unpopular. - **Z7 — Ideas interview vs accountability interview `#atom/zettel/z1`.** Format determines which questions are askable. Reputation effects can be a byproduct of format rather than intent — the key distinction the ChatGPT reading collapses. ### Fleeting Queue - `#atom/fleeting/f1` The only use of "governance" in a transcript about steering an ecosystem refers to a city council. Worth a short post on its own. - `#atom/fleeting/f1` Security and performance are named as the durable priorities and then given one mention each — the episode enacts the neglect it diagnoses. - `#atom/fleeting/f1` Does Crossword's Matt episode + this Josepha episode constitute a deliberate two-part governance arc? Check the S11 run order. - `#atom/fleeting/f1` "You only see the final mile, which was already the hardest, most expensive mile" — reusable for the publishing-friction argument in Field Notes. - `#atom/fleeting/f1` Verification note for the pipeline: absence claims need `\b` anchors. `matt`→"matter", `mary`→"summary" would have produced a false refutation. Consider folding into the deep-read skill. <!-- deep-read:sections end --> 📄 **Full text:** [[Full Document Contents/Articles/Perspectives With Josepha Haden Chomphosy.md|Perspectives With Josepha Haden Chomphosy by Crossword]] ## Highlights - **Josepha:** I’ve heard. I’ve heard. Yeah, I think that the three things that are always predictable for the future of WordPress are that people who are building websites or building apps on top of WordPress would like it to be more secure — both from a core perspective and everything that you can bolt onto core, like both of those need to be secure. I think they always want it to be fast — sometimes not faster, just not any slower is gonna count as well. And then also easy to use. And that’s not in an “I have my scissors and I know how to use my scissors” way, but easy as in: I know exactly what I wanted to do, I was able to figure out the gaps to what would make that possible, and then it was easy for me to get it done. And the “get it done” part is negotiable. Like, you personally can do it, you can vibe code it using your favorite GPT, you can pay somebody — it doesn’t matter. The question is, from an ecosystem perspective, again: is it easy for people to tell what comes in the basic bits? Is it easy to tell how far away their preferred future is? And is it easy for them to get there? That’s everything that everyone ever wants. ([View Highlight](https://read.readwise.io/read/01kze4ap58xx3e393r4k6k8k3f)) - Note: L1/L2 — The three predictables (secure / not-slower / easy) as a prioritization invariant, plus her journey-based definition of "easy": know what's in the basic bits, know how far your preferred future is, know how to get there. The most portable artifact in the episode; usable directly as a feature gate. - Tags: wordpress, atom/zettel/z1, prioritization, atom/literature/h1 - **Jonathan:** I don’t know — as a constituent I think it’s an important moment for this ecosystem. I’m noticing, and Luke continues to give me a hard time about it, that I’m doing less with WordPress and more with just static sites. I’m taking things off of WordPress. We’ve been talking over the arc of this season about how it feels temporary, and there’s this part of me that’s like, I’m expecting that WordPress is going to get easier to use. I don’t generally think that much about the security side of it, for better or for worse — I’ve sort of taken for granted that it’s pretty secure. And the performance I do care about. But we’re in this moment right now, as I’m listening and thinking about it, that I’m not sure that those — I don’t see explicit evidence that those remain key priorities. Right now we’re trying to figure out where does AI fit into this stuff, and are we making that, I suppose, easy to use? But it’s unclear to me. And when I think about it, I see this tension between — in a proprietary ecosystem, you’re either paying attention to these things and trying to guide and influence them or you’re not. And then in open source it’s like, well, who is doing that? And how do you measure whether that’s being effective or not? What’s interesting to work on is kind of what gets the attention, and we’re not necessarily connecting that to how it’s impacting the ecosystem. There’s a lot of good work that happens. It’s just hard to know — at least as an observer, if you’re not in it all the time — what direction is this taking us? So yeah, as someone who has a vested interest from a personal and professional perspective in the WordPress space, it’s a bit painful to me that I don’t know the answer to that. There’s a lot of activity. There’s a lot of things that are happening. But I’m still sort of waiting to see where this is going to go. ([View Highlight](https://read.readwise.io/read/01kze4d6frfr6marhtm9ebn7rn)) - Note: L3 — The core diagnosis, and the episode's #1 problem: "What's interesting to work on is kind of what gets the attention, and we're not necessarily connecting that to how it's impacting the ecosystem." Plus the accountability vacuum stated plainly: "in open source it's like, well, who is doing that? And how do you measure whether that's being effective or not?" Note the fallacy tag: his static-sites migration is n=1 personal behavior functioning rhetorically as ecosystem evidence — he's honest about it, but it still does argumentative work it can't support. - Tags: prioritization, fallacy/anecdotal-evidence, atom/zettel/z1, governance, atom/literature/h2 - **Luke:** And I was hoping to get through the podcast without mentioning AI, but you did it first. So, adding on to that — it seems like the current big push, the current big focus within WordPress, the current interesting thing to work on, as you put it, Jonathan, is AI tooling. And I’d suggest that these AI features that we’re working on are making WordPress less secure, no faster, and harder to use. ([View Highlight](https://read.readwise.io/read/01kze4at39x8cpxwf5twyefcx8)) - Note: L7 — Negates all three predictables in one line. Most quotable claim in the episode and the least supported: no feature named, no measurement offered. Also the most testable — Luke sits on Plugin Review and his team holds the queue telemetry that would settle it. - Tags: ai-strategy, atom/literature/h3, fallacy/hasty-generalization, wordpress - **Josepha:** Gosh, what were the public toilets when I first brought that up? I can’t remember what it was, but I do remember that I met a bunch, all of a sudden, of long-tenured old WordPress people at the time — this was ten years ago — and they’re like, now listen, you. And I was like, yeah, I get it. I would hate me too. But also, I think we still have to do it. ([View Highlight](https://read.readwise.io/read/01kze4axje1sns997zgg02y17r)) - Note: The missed follow-up. Luke asks what WordPress's neglected infrastructure actually was during her tenure; she can't recall, and nobody pursues it. This is the natural door to "what did you prioritize, and did it work?" — the accountability question the episode never asks. Note the shape of the self-criticism: "I would hate me too" is self-implicating in form but self-validating in substance (I was right and unpopular). Neither "regret" nor "mistake" appears anywhere in the transcript. - Tags: open-source-leadership, atom/literature/h2, governance - **Josepha:** I’m gonna answer an underlying part of this question first that applies to both of these contexts, which is that open source and cities are currently experiencing this issue with being able to get information to people, but also being able to get information back from them. It is so hard to hear the people that you’re working for right now. And that’s true in open source. While I was the ED, so much of my time was begging people for opinions, and begging people for information — clear information that went with their feelings — and gathering folks together to just listen to them talk to each other. Because they’re more honest with each other than they are with, in my case now, their city government, but at the time, with me. My very presence made people slightly less honest. And so when we look at a centralized thing that is happening to something that was meant to be decentralized — on the one hand, no one likes this answer, and WordPress didn’t like it when I started doing it either — but if you want to get much further than you were when I got there, you have to have some centralized resources, just to make it all make sense. And also for efficiency’s sake, which is uncomfortable. ([View Highlight](https://read.readwise.io/read/01kze4bqz9jrdm6nvx64ngn4fm)) - Note: L4/L5 — Two atoms in one paragraph. (1) Presence-distorted feedback: "my very presence made people slightly less honest" is the single most genuinely self-implicating line in the episode — an admission that her leadership method had a structural defect. Corollary: convene people to talk to each other, not to you. (2) The centralization concession: some authoritative shared infrastructure is a precondition for a working decentralized ecosystem. Politically significant coming from a former ED; directly relevant to FAIR's federated-vs-central design debate. Also the only place the ED role surfaces at all — she never says "executive director." - Tags: atom/zettel/z1, open-source-leadership, atom/literature/h2, governance - **Josepha:** And it’s super uncomfortable, especially for open source. It’s really uncomfortable for people who find themselves in a democracy — because it’s not a direct democracy in a lot of cases. But in our city, we would absolutely love it if people would show up to our town hall meetings and just be like, I don’t love that my taxes go to parks, and then we could figure that out, right? I was desperate half the time in WordPress for someone to show up and not only be like, I hate this thing, but then say: because of this. Just tell me the because. If you say that everyone you’re talking to is hating that thing because of the color of that button, tell me how many everyones, and tell me what the better button might be, and I’ll see if we can make those things work out. But the decentralized nature of things is also, in a situation like this, entropy — which can be a catastrophic state for a large organization. I don’t know much about the movement of information in WordPress at the moment, but I do know that if you’re looking at 43% of the web, that is a ridiculously huge number of people that are actively engaged in the thing that you are working on, and passively experiencing the thing that you’re working on. And our next frontier was getting to those passive people, because we had made a tent for them and still couldn’t get them. And so the question of how to keep the decentralized value — the value proposition that that brings — while also making it easier to accomplish tasks on behalf of those opinions, is a really, really complicated proposition. It is plaguing many open source communities, and certainly it is showing up in local government as well. ([View Highlight](https://read.readwise.io/read/01kze4ceezjdh9jzq7qgrpg91h)) - Note: L6 — The passive majority as unreached frontier: "we had made a tent for them and still couldn't get them." Openness of participation does not produce representativeness of participation. Paired with the actionable feedback ask — "just tell me the because": magnitude, affected population, proposed alternative, instead of sentiment. Build the intake form around this. - Tags: wordpress, atom/literature/h3, governance