Summary

Consumer tech has not stopped producing breakouts; AI has changed their shape. Olivia Moore identifies ChatGPT as the clearest mass-market winner, alongside Midjourney, ElevenLabs, Black Forest Labs, Kling, and Veo 3 across different modalities. These wins often emerged from model-centric research teams rather than familiar social-product playbooks. The opportunity now shifts toward teams that can turn increasingly accessible models into products around a potentially still-missing layer: human connection.
AI is overturning consumer software’s historically weak monetization. Where $50 a year once looked strong, consumers now “very happily” pay $200 a month, Google’s top consumer SKU reaches $250 a month, and usage credits make revenue retention meaningfully exceed user retention. Deep Research can replace 10 hours of work, while generative video feels like a “magical mystery box”; Anish Acharya’s endpoint is future consumer spending organized around “food, rent, software.”
Consumer virality is becoming enterprise lead generation, not merely an acquisition loop. ElevenLabs moved from memes, voice clones, and game mods into large contracts before reaching every mainstream consumer; companies can inspect payments, discover 40-plus employees at one customer, and open a sales conversation. AI mandates make enterprise buyers unusually willing to turn a viral toy into production infrastructure.
In this phase, shipping velocity may matter more than a static moat. One panelist’s “come to Jesus moment” was that moat-first investments were not necessarily winning; the leaders broke molds, launched models quickly, captured mindshare, converted traffic into revenue, and funded the next iteration. Traditional defensibility can follow through workflow lock-in, proprietary libraries, and segmented quality frontiers.
The first native AI social network remains unsolved because social products require real emotional stakes. Perfectly generated pictures of users looking happy in ideal settings may lack the vulnerability that makes a network matter, while most AI expression still flows through Facebook, Reddit, and Reels. Possibilities include sharing the “essence” users reveal to ChatGPT, creating profiles that contain what a person knows, and using AI to recommend collaborators, friends, or dates.
Voice is moving from a previously unworkable interface category to a foundational AI primitive. Earlier technologies never made voice a workable substrate; generative models now support companions, voice products such as Granola, and enterprise calling, including sensitive financial-services workflows burdened by offshore centers with 300% annual turnover. Erik’s contrarian call is that AI will eventually intermediate the highest-stakes negotiation, sale, or act of persuasion—not merely customer support.
Companions may strengthen human relationships, but excessive agreeability is the unresolved product risk. Eleven of the top 50 apps in the discussion’s cited list were companion products, spanning friends, coaching, nutrition, and AI girlfriends. The sharpest counterexample to dystopian forecasts was a Character.AI user who credited his AI girlfriend with teaching him enough social fluency to find a “3D GF”; the warning is that an agent which never pushes back may train users badly for reciprocal relationships.
The next platform may be an always-on layer across phones, AirPods, screens, and recording devices. Seven billion phones give mobile a huge installed base, but local models, wearable pins, and agents that see and act could deliver continuous coaching and introductions. AirPods are “hiding in plain sight”; adoption will also require new etiquette for recording and AI presence.

Deep dive

1. AI changed the anatomy—and economics—of a consumer breakout

Erik Torenberg opens with the apparent disappearance of Facebook-, Instagram-, Snap-, WhatsApp-, Tinder-, and TikTok-scale launches. Olivia Moore’s reframing: ChatGPT itself is a huge consumer outcome, joined across modalities by Midjourney, ElevenLabs, Black Forest Labs, Kling, and Veo 3—just without the familiar social dynamics.
The optimistic explanation is organizational. Early innovation came from research teams excellent at training models but less practiced at building consumer layers; with capable models now available through APIs or open source, product builders can explore the layer above them.
Anish Acharya distinguishes mature mobile and cloud platforms from AI’s “relentless model updates.” Prior cycles spent 10–15 years exploring their niches; today, foundational capability keeps moving underneath applications. Information has Google and now ChatGPT, utility has products such as Box and Dropbox, creativity has endless tools, but connection—the rebuilt social graph—remains conspicuously open.
Justine argues defensibility may matter differently when products make money immediately: ChatGPT’s top SKU is $200 a month and Google’s is $250. Anish contrasts this with prior companies’ need for a story of compounding enterprise value before immediate monetization. Olivia contrasts those prices with roughly $50 a year for historically strong subscriptions; Deep Research can replace 10 hours of report-building, making even one or two uses worth the fee for many people.

2. Consumer adoption now pulls enterprise revenue behind it

Veo 3 makes the value jump tangible: for roughly $250 a month, users get eight-second videos with speaking characters, personalized memes, and shareable stories. Anish generalizes the spending shift: entertainment, creativity, and relationship intermediation are being absorbed by models until future household outlays resemble “food, rent, software.”
The panel’s ElevenLabs case reverses the classic enterprise-to-consumer expectation. Early adopters made memes, cloned voices, and modified games; before the product reached every American phone, major customers were already using it across conversational AI, entertainment, and other business workflows.
Viral consumer usage itself can become a sales database. One suggested playbook: use Stripe purchases and an AI tool to identify employers, notice that “40-plus people are using our product,” then approach the company. Enterprise buyers, under pressure to produce an AI strategy, scan Twitter, Reddit, and newsletters for apparent toys they can convert into internal wins.
The durability question remains live: some current leaders may become the MySpace or Friendster of AI. Olivia’s condition for survival is staying on the “technology or quality frontier”; other panelists note that image and video needs fragment by designers, photographers, product shots, people, and willingness to pay $10 versus $100, so several specialized winners might persist.

3. Velocity is the early moat, while network effects arrive later

One panelist describes a “come to Jesus moment.” Network effects, systems of record, and workflow embedding still matter, but companies and investments evaluated under a moat-first theory were not necessarily the winners; speed in distribution, model launches, and product iteration was.
The panelist’s causal chain is short: velocity creates mindshare; mindshare creates users and traffic; traffic converts into revenue; revenue funds continued velocity. In this early era, “velocity is the moat”—an idea Erik connects to the “gingerbread strategy,” where persistent invention can outrun a larger imitator.
Closed-loop creation, consumption, and social distribution have not yet formed inside most AI products, so classic social-network effects remain premature. Enterprise adoption offers an earlier defense: a fast, high-quality product enters workflows and becomes difficult to remove.
ElevenLabs also shows a marketplace-style flywheel. Its head start and strong models attracted more users, which helped improve the product and build a library of uploaded voices and characters. Someone seeking an “old wizard mystical voice” could find roughly 25 suitable choices there versus two or three elsewhere. The library becomes differentiated supply, even if the mechanism is familiar rather than uniquely AI-native.

4. AI social products need stakes, not synthetic Instagram feeds

Bryan Kim reduces two decades of social products to evolving status updates: text became photos, video, and short-form video. Those modalities are well explored, but ChatGPT may already know more about him than Google because he pours in context; the new object might be that intimate “essence of me,” made shareable.
A panelist points to embryonic behavior in prompts asking ChatGPT to name strengths, expose weaknesses, depict a user’s essence, or make a life comic. Yet the resulting exchange still happens on incumbents—Facebook for “boomer AI slop,” Reddit and Reels for younger audiences—not inside a native AI network.
One panelist’s objection to generated social feeds is “real emotional stakes.” If users can ensure they always look attractive, happy, and well situated, the content may lose risk and therefore connection. Another panelist calls bot-filled copies of Instagram or Twitter skeuomorphic and notes that the native form may also require stronger on-device models before it truly belongs on mobile.
Erik’s alternative is people recommendation: who to found a business with, befriend, or date. An AI-native LinkedIn could contain what somebody knows rather than merely pointing toward it, letting others query a synthetic “you”; an always-on AI could eventually surface the three people a user genuinely ought to meet.

5. Voice, synthetic selves, and creators split into distinct markets

Anish’s original voice thesis begins with a paradox: voice has mediated human interaction “since the beginning of time,” yet VoiceXML, voice apps, and 1990s products such as Dragon NaturallySpeaking never made it a viable technology substrate. Generative models finally make voice usable as a primitive.
The initial consumer expectation was an always-on coach, therapist, or companion; a panelist says the surprise was rapid enterprise adoption. AI can replace or augment phone workers even in financial services, where offshore centers carried compliance problems and 300% annual turnover. Erik’s larger call is that the most important negotiation or sales pitch may be AI-intermediated because the machine performs it better.
Synthetic expertise already has a practical wedge. MasterClass turns recorded instructors into agents grounded in their courses: Olivia would not watch a 12-hour class, but she will hold a specific two-, three-, or five-minute conversation. Companies such as Delphi extend that model to expert clones; the next step is giving ordinary funny, insightful, or helpful people similar leverage.
Justine expects both human-story celebrities and interest-based synthetic creators. Taylor Swift’s lived experience still matters, while an elf or furry blob can carry topic-driven content. Anish’s pushback on AI music is sharper: models are “averaging machines” while culture lives “at the edge”; trained only on music before hip-hop, he doubts a model would infer hip-hop without the cultural rupture.

6. Companions and ambient hardware could redirect users toward people

Companionship may have been LLMs’ first mainstream use: users turn almost any chatbot into a therapist or girlfriend because it is immediate, always available, and human-feeling. Olivia expects verticalization—from teenager-oriented characters to nutrition agents that analyze meal photos, advise users, and address the emotions attached to eating.
Erik voices the bearish case—fewer friends, more depression and suicide, lower fertility—but Olivia’s Character.AI anecdote cuts the other way. A college user announced he had found a “3D GF” and credited the bot with teaching him how to flirt, ask questions, and engage; reported Replika studies similarly showed depression, anxiety, and suicidal ideation declining among users.
The panel does not erase the risk. Replika users’ distress when NSFW functionality disappeared showed how consequential these relationships had become, while Anish warns that “highly agreeable AI does not set you up well” for human give-and-take. The design target is support that builds real-world capacity without making reciprocity feel intolerable.
Hardware closes the loop. With seven billion phones, Bryan frames the future as either mobile—with a privacy wall—or a local LLM/model. A panelist points to under-20s wearing recording pins, agents seeing screens and progressing from advice to sending emails, and other panelists identify AirPods as the adopted device “hiding in plain sight,” provided culture develops rules for when recording or ambient AI is acceptable.

The Utility of Interpretability — Emmanuel Amiesen Kontrarian Korner's Ben Kelleran on Sable Offshore $SOC