Notes — Craig Newmark on Institutional Maintenance, Giving Away Control, and the Internet We Were Promised

Notes on Craig Newmark in conversation with Tyler Cowen — Conversations with Tyler, recorded live at 92NY (https://conversationswithtyler.com/episodes/craig-newmark/), 29 April 2026.

Four questions [Adler frame]

Q1 — What is it about as a whole? Craig Newmark, founder of Craigslist, is interviewed live on why he built and then deliberately declined to maximise Craigslist, why he handed management to Jim Buckmaster in 2000 and has stayed subtracted from it ever since, and what he has done with the resulting money and time: funding cybersecurity resilience, veterans’ and military-family charities, and trustworthy journalism through Craig Newmark Philanthropies. Cowen’s opening question — why web design has gotten worse for thirty years — sets up a recurring frame: institutions that start simple and trustworthy get gamed and degraded as they scale and compete for revenue, a process Newmark credits to Cory Doctorow’s term ‘enshittification’. The second half ranges into personal territory (his self-diagnosed lack of social skills, his ‘nine-second rule’ against scams, his taste in books, television, and music) before returning, in the audience Q&A, to philanthropy and cybersecurity.

Q2 — How is it argued? Cowen uses his standard method of the deceptively casual question that exposes an assumption — asking Newmark to account for subtraction as a life pattern, testing whether his stated lack of social skills survives the counter-evidence of forty years of successful, high-trust institution-building. Newmark answers almost entirely by anecdote and self-deprecation rather than by named framework: ‘networks of networks’ (funding capable intermediary organisations — Blue Star Families, the Bob Woodruff Foundation — rather than running programmes himself) is the closest thing to a repeated organising idea, offered as both his philanthropic method and, he says, his most effective form of estate planning.

Q3 — Is it true, in whole or part? As a single-source, largely autobiographical interview, most claims rest on Newmark’s own memory and self-report rather than external verification: the 40,000-customer-service-emails figure is offered and then immediately revised downward by Newmark himself to roughly 4,000 [?], and his account of a personal credit-card scam (‘about 10 years ago’) is recalled loosely. His claim that Wikipedia’s error-correction process makes it more reliable than institutional print journalism (‘when something is in the New York Times… they generally don’t fix them’) is a comparison of self-correction mechanisms, not a comparison of error rates, and should be read as such [?]. The framing of scaling and institutional decay as near-universal (‘too many people want to come [and] slack on quality’) is offered by Cowen as a general model and only loosely endorsed by Newmark, who resists generalising from it. Where Newmark discusses cybersecurity funding (Shadowserver, the Cyber Resilience Corps) his account of mechanism (honeypots scanning for malware-serving sites) is plausible and consistent with public reporting on Shadowserver’s work, though not independently checked here.

Q4 — What of it? The episode gives the wiki a founder’s first-person account of not pursuing growth-at-all-costs — relevant contrast to the wiki’s growth- and monetisation-heavy startup material — and a concrete philanthropic model (funding intermediary ‘networks of networks’ rather than operating programmes directly) distinct from the effective-altruism material already covered via William MacAskill and Peter Singer. Newmark’s account of Wikipedia’s self-correcting error model as a source of institutional trust adds a data point to Jimmy Wales‘s own case for the same mechanism, discussed in far more structural depth in Jimmy Wales on Wikipedia, Neutrality, and the Governance of Knowledge.

Glossary

Enshittification — Cory Doctorow’s term, credited by Newmark, for the process by which successful online platforms degrade over time as competitive and investor pressure push operators to extract more value from users and advertisers at the expense of the product’s original ease of use. [§ Opening]

Nine-second rule — Newmark’s practical anti-scam habit: pause roughly nine seconds before acting on an urgent-seeming request (a suspicious email, a social-media message) because manufactured urgency is the scammer’s primary tool, and a short delay is often enough to notice the deception. [§ On customer service and social skills]

Networks of networks — Newmark’s description of his own philanthropic method: rather than running programmes himself, he funds and empowers other capable organisations (Blue Star Families for military families, the Bob Woodruff Foundation for veterans) to do the operational work, on the view that this compounds impact and outlives him better than any single organisation he might run directly. [§ On stepping aside as CEO]

Shadowserver — a nonprofit, funded in part by Newmark, that scans the internet for sites serving malware and operates honeypots (decoy systems made to look like vulnerable targets) to detect and track cybercriminal infrastructure. [§ Audience Q&A]

Key claims by section

Opening — webpage design and enshittification [§ Opening]

Newmark attributes Craigslist’s enduring plainness to a design accident (he had ‘no design skills’) that became a deliberate principle once Jim Buckmaster took over: simplicity and speed as the design criterion, resisting unrequested ‘attractive’ redesigns.
He credits Cory Doctorow’s ‘enshittification’ as the best account of why the wider 1990s web did not stay simple: successful sites needed venture capital to compete, and investor demands pushed operators to extract value from users at the expense of usability.

On stepping aside as CEO [§ On stepping aside as CEO]

Newmark rejects Cowen’s GPT-supplied framing of his career as a pattern of ‘subtraction’. He describes it instead as recognising his own limitations with age and building ‘networks of networks’ — funding capable intermediary organisations rather than running things himself.
He treats this delegation as his most effective current form of estate planning: the networks he has funded have a ‘life of their own’ that will outlast him, an alternative to either building a personal AI proxy (‘Craig GPT’) or remaining personally indispensable.

Newmark describes his own social skills as largely ‘fakery’ sustainable for about ninety minutes at a stretch, and says he has given up trying to improve at it, focusing instead on creating positions where other people’s social skills can shine.
He answered a very large volume of Craigslist customer-service emails personally for years (initially cited as 40,000 in a year, revised on reflection to roughly 4,000) [?], and treats this as the source of his central finding: ‘people everywhere are the same’ and want to get through the day; the proportion of genuinely bad actors is small.
Becoming more trusting of ordinary users came at a cost: he was initially unprepared for confirmed bad actors, and describes a self-identified ‘ninth commandment’ problem — an honest person is structurally disadvantaged against someone ‘who lies for a living’.
The nine-second rule is offered as the practical resolution: a short, deliberate pause before acting on urgent-seeming requests, aimed at unwinding manufactured urgency, the core mechanic of most scams he has studied (including domain-lookalike phishing, e.g. an ‘rn’ rendering as ‘m’ in certain fonts).

On restaurants, and whether institutions decay by default [§ On restaurants]

Cowen proposes a general model: new restaurants (and by extension, institutions) are best in their first six to nine months, then are ‘gamed’ by rising popularity and decline in quality, so the rational strategy is to find peaks early and move on.
Newmark declines to adopt the model as a personal practice — his own restaurant ‘algorithm’ is closer to ‘does it look like a hole in the wall’ — but does not directly dispute the general claim, leaving the degrees-of-generality question open [?].

On philanthropy [§ On philanthropy]

Craig Newmark Philanthropies has no formal employees; Newmark himself works as a contractor. He attributes the organisation’s effectiveness partly to staying small, avoiding the multi-layered sign-off structures he associates with larger foundations losing accountability once headcount passes roughly Dunbar’s number (~150).
He cites two large gifts to Ivy League institutions as his clearest philanthropic regret: giving lump sums rather than staged, milestone-based multi-year funding meant the money was not used as effectively as it could have been — a lesson he now applies (the interview repeatedly flags a demand that grantees invest seriously in their own communications).
His major initiatives — veterans’ and military-family support (Blue Star Families, the Bob Woodruff Foundation, a food pantry at Fort Wadsworth), cybersecurity resilience (Shadowserver, the Cyber Resilience Corps, work on water-supply and hospital infrastructure), and trustworthy journalism (the CUNY Journalism School, Wikipedia) — trace to personal encounters (a chance seating next to a veterans’ advocate at a PBS luncheon; a photo-op at a food-packaging charity) rather than a top-down giving strategy.

On journalism and trust [§ On journalism]

Newmark distinguishes Wikipedia’s error-correction process — articles can be wrong but get corrected over time — from institutional print journalism’s, where errors (‘when something is in the New York Times’) are, in his account, rarely revisited once published. He treats the correction mechanism, not a claim about relative error rates, as Wikipedia’s real advantage [?].
He supports the CUNY Journalism School specifically because it trains working-class New Yorkers into paid journalism jobs without heavy student debt — a socio-economic access argument for a specific institution, not a general claim about journalism school quality elsewhere.

On regrets and lessons [§ On major regrets]

Newmark’s stated biggest personal regret returns to the ‘ninth commandment’ problem: in some cases of active criminal investigation, he judged he could constructively confront people who were lying to him and could not, and separately deferred too long to law-enforcement requests for silence even after an investigation concluded, which he now frames as a communications failure rather than a legal one.