Substack Cofounder on AI Slop Content & the Decline of Social Media
Summary
Substack’s core bet is that independent creators need both economic independence and an incentive-aligned distribution network. Chris Best calls free speech a necessary precondition, but the larger mission is a “new economic engine for culture” in which creators make money, retain editorial control, and connect directly with audiences.
The right to exit became a moat because it forces Substack to earn creator loyalty rather than manufacture lock-in. Writers can export their email lists, while returning publishers become beloved “boomerangs.” More importantly, subscriptions let creators “override the algorithm,” call in audience trust, and take risks that engagement-optimized platforms would bury.
Substack’s broader network vision addresses a structural dependency that paid newsletters alone could not solve. In 2018–2019, successful writers still needed Twitter, Facebook, or LinkedIn as top-of-funnel infrastructure; any platform policy change could become an “existential event.” Substack therefore wants a network built on “different laws of physics,” where discovery can lead to deep engagement and payment rather than more screen time.
The episode treats algorithms, advertising, and AI as tools whose outcomes depend on their objective functions and incentives. Best warns that copying legacy ad systems would import the same conflict between platform and participant, while well-designed sponsorship tools might expand creator income. Torenberg describes AI that could generate engagement slop or give independent voices enough production leverage to turn a FaceTime-like conversation into video, podcasts, clips, transcripts, and multiple languages.
Attention—not content—is now the scarce input, making trusted curation and high-quality creator brands more valuable. Best argues society has “won the war on boredom”: nobody lacks something to watch, but worthwhile material remains scarce. The consumer proposition is therefore spending money to acquire “better culture, better ideas” because media is both time spent and “who you become.”
Substack’s creator economics could produce the same talent migration and positive variance that venture capital produced in software. The episode contrasts a hypothetical $80,000 institutional salary for Noah Smith with “a million dollars or whatever it is” independently, then extends the model from solo creators to Substack-first media companies. Best’s ambition is to put “the lunatics in charge of the asylum” and give ambitious media founders infrastructure for larger institutions.
The $100 million round funds Substack’s transition from newsletter platform to scaled cultural network. Best says the “fledgling network” is now alive and growing; capital will help rebuild the product and company around a handshake between creator independence and an internet-scale destination. Best frames the broader media future as a choice between increasingly potent “AI goon bots” and media people are grateful to have consumed.
Deep dive
1. Independence made free speech economically durable
Katherine Boyle reconstructs the summer of 2020 as a fearful rupture: New York Times opinion editor James Bennett was forced to resign after publishing an op-ed by a sitting senator; there were mass firings; Twitter deplatformed a sitting president; and people feared that asking heterodox questions could cost them their livelihoods. Against that backdrop, “one platform stood up and said, ‘Hey, we are protecting free speech,’ and that was Substack.”
Boyle’s claim is deliberately counterfactual: today’s wider Overton window makes the earlier pressure easy to memory-hole, especially after Elon bought Twitter in November 2022. She argues the culture “could have been living in a totally different time” without infrastructure willing to support displaced or dissenting writers.
Best narrows the framing: free speech has always been “an important pillar, but not the main pillar.” Substack’s primary project is a “new economic engine for culture,” grounded in the belief that great work comes from independent voices with editorial freedom, meaningful income, and a direct audience relationship; freedom of speech and press are necessary preconditions for that system.
His surprising observation from 2020 is that the people feeling the sharpest pressure were not necessarily conservatives or Republicans, but often “the best and most interesting people” inside liberal media. Their removal from established institutions aligned unusually well with Substack’s business model: “It was spicy from a cultural perspective, but that’s the gig.”
2. “Blogging with a business model” was the narrow wedge
Andrew Chen remembers an internet where LiveJournal, Zanga, Blogger, and Google Reader had faded, while the open WordPress ecosystem offered weak economics, spam, hacked PHP sites, affiliate links, and AdSense. Ben Thompson’s Strateery suggested another model, but assembling publishing, payments, and operations remained cumbersome. Best welcomed the early accusation that “Substack is just blogging with a business model”: “That sounds pretty good.”
Best traces the company to an essay he began in 2017 while trying to become a writer. Hamish, his writer friend and future cofounder, gently rejected another complaint about newspapers and Facebook: if the diagnosis was already understood, “what could you do about it?” Their answer was techno-optimist rather than nostalgic—put new technologies and networks “in service of people” and actively steer toward the better future.
The grand vision started with a tiny, concrete product: make paid email newsletters dead simple. Best estimates perhaps 20 people initially “really, really wanted it,” but for them it would be transformative; even the first newsletter worked without a pre-existing network, offering a way around the cold-start problem. YouTube was the nearest analogy for the eventual ambition, though Substack began with the smallest viable expression of it.
3. Portability and discovery form one creator contract
Substack initially intended to allow only paid publications because payment aligned incentives. Its first customer immediately punctured that purity by proposing Mailchimp for free readers and Substack for paid ones. Best concluded that a successful paid publication required a free funnel, and supporting both was not “an abrogation of our vision” but a practical condition for realizing it.
The same dependency reasoning led to Substack’s wider-network plan. In 2018–2019, an independent publisher still needed Twitter, Facebook, or LinkedIn for discovery, leaving the business downstream from platforms with no intrinsic interest in creator income. If Mark Zuckerberg decided to suppress politics, Best notes, a politics creator dependent on Facebook faced an “existential event.”
Email ownership supplies both exit and creative agency. Best says allowing exports appeared foolish to observers who asked whether customers would simply leave, but it forced Substack to create enough value that publishers stayed—or returned as “boomerangs.” A subscription is also permission to “reach out and tap you on the shoulder,” enabling creators to spend audience trust on work an engagement algorithm might never distribute.
Best observes that algorithmic “For You” feeds increasingly disregard the follower graph: 100,000 followers may see none of a creator’s work. Torenberg rejects an anti-algorithm stance: algorithms are powerful tools serving specified ends. Substack’s opportunity is an objective function whose “secret hidden master” is the reader’s deeper interest, rather than advertising volume.
4. Ads and AI are leverage only if incentives survive
Asked whether video and valuable audiences imply an ad network, Best draws a boundary: copying legacy social advertising would import the business model that puts platforms “at odds with the people on the platform.” Yet many excellent Substack writers already sell sponsorships, so a first-principles system could unlock additional income while preserving independence and differentiated quality.
Torenberg retains the trade-off: advertising can reward clickbait and content “dumbed down” for scale, but it can also let niche writers monetize without charging readers heavily and may reduce subscription-driven audience capture. Best does not offer a finished mechanism; the requirement is to discover “the good version” rather than merely stuffing ads into the product.
The conversation applies the same fork to AI. Torenberg describes a live product resembling FaceTime that can become a produced podcast, YouTube video, short clips, transcript, and eventually any language. He argues this production flywheel matters because “the hardest part about writing is writing”; his split is between “AI slop” engineered to keep people clicking and creative leverage that helps independent voices realize more ambitious work.
5. Scarce attention creates both a quality premium and a hazard
Best dates the decisive scarcity shift to the internet and social-media era, not generative AI. Early platforms entered a land grab where consumers had spare attention and wanted free distraction; now society has “won the war on boredom.” Content is abundant, but attention remains finite and there is “a huge scarcity of things that are worth paying attention to.”
That inversion underwrites subscriptions: for someone with one life, paying for better ideas, culture, and uses of time is “a phenomenal deal.” Best’s deeper proposition is that media “is not just how you spend your time, it’s who you become”—so selection affects individual character, cultural norms, and the value attached to trusted creators.
His “two futures of media” separates immediate stimulation from culture. One path approaches science-fiction wireheading: increasingly sophisticated “AI goon bots” deliver compelling pleasure while degrading taste and pulling against long-term interests. Best treats this path as “baked in” and accelerating, not as a possibility Substack can eliminate.
The contingent alternative is media that remains fun while helping people understand society, participate, and become who they aspire to be. Best does not predict universal conversion—“Some people are going to sit on the AI goonbot”—but wants users to leave Substack thinking, “Damn, I’m glad I did that. That made me a better person.”
6. Creator economics can reorganize institutions beyond media
Torenberg frames Substack as price discovery for talent: someone like Noah Smith might be valued at roughly $80,000 inside Bloomberg but “a million dollars or whatever it is” independently. Best recalls Mark Andrees’s analogy to venture capital freeing software builders from suited managers: putting makers in charge increased variance, failed sometimes, but produced far more exceptional outcomes.
Best applies that model to culture: creators “are the heroes,” and putting “the lunatics in charge of the asylum” could generate a renaissance. The endpoint is not exclusively solo publishing; Substack has a team dedicated to making the platform the best place for “ambitious media founders” to build larger operations and realize the biggest version of their ambitions.
On academia, Best calls himself “a total crank” and says much of science, especially academic publishing, is broken; he even thinks peer review “maybe” was a huge mistake. His tentative alternative is radical in its simplicity: let researchers publish science directly online. He sees early shoots, though the category has not been central to Substack’s effort.
Best rejects the recurring panic that internet writing without editors or reading outside books means people have stopped reading; he sees it as legacy industries adapting. Torenberg, citing Alex Danco, notes that long-form writing matters partly because influential readers translate it into other formats, while Chen adds that internet discourse moves faster and long-form generates original thoughts. Chen contrasts the roughly three-year process of writing a book, a bestseller threshold around 10,000 units, and what he says may be a single remaining US book printer with instantly reaching 100,000 inboxes. The new $100 million round is meant to scale that entire network-and-independence model.