Summary

Kuyda sees today’s chatbots as the “MS-DOS era for AI interfaces”: enormous adoption has validated demand without exposing the models’ full capability. Almost a billion people use AI tools, yet search, homework, and writing—roughly one-third of usage in the research she cites—dominate because a command line advertises commands. The platform opportunity is a Windows/macOS-style visual layer that makes advanced use cases ordinary.
Personal software makes tiny, temporary, deeply customized apps viable where App Store economics cannot. Kuyda’s examples include a two-minute Elsa-and-Jasmine puzzle translated into Italian, a 5:30 a.m. quote app drawing from one television show, and a lifting tracker continually adapted to her book, gym, and goals. The destination is an operating system “built on the platform of you.”
Wabi’s ambition depends less on turning everyone into a coder than on becoming the organizational layer for software made by everyone. Kuyda expects fully original creators to remain below 10%, with many more people remixing, requesting changes, or adjusting styles. Mobile distribution, guardrails, integrations, social discovery, and shared context and memory provide the platform value that loose vibe-coded links lack.
Apps could become executable media: creator content, monetizable protocols, and community starters rather than fixed utilities. A fitness influencer might distribute five mini apps instead of a course; a designer might publish a distinctive Pomodoro timer; multiplayer apps could organize dog owners or neighborhood parents. Software creation is the remaining “last frontier” for creators.
The strongest platform value in Kuyda’s account is cumulative context shared across mini apps, not any single generated interface. Features, appearance, prompts, connected services, personal goals, and platform memory can all adapt; eventually a nutrition app might inherit relevant context from a workout app. Her thesis is “deep, deep, deep” personalization rather than old software wearing an AI wrapper.
Replika’s history is a warning that recognizing a wave early is insufficient when capital and execution do not match the opportunity. The company bet on generative dialogue in 2015, discovered the breakthrough was seven years away, and later prioritized revenue after raising only $1 million rather than pursuing a contemplated $20 million model-building push. Kuyda does not claim that bet would have succeeded, but her lesson is categorical: sometimes founders must “go big or go home.”
For hardware, Kuyda rejects the “huge mind trap” that voice should be the primary AI interface and argues for a screen-first operating system. Voice fails in bed, offices, crowds, while walking, discovery, and rapid information intake; she notes that roughly 75% of Alexa devices ship with screens. Her preferred future combines local models, fluid software, and persistent personalization: “AI is just an app on your phone. It should not be that way.”

Deep dive

1. Chatbots are the MS-DOS layer, not the destination

Kuyda locates one through-line from Replika to Wabi: machines should understand language and use that understanding to improve an individual’s life. AI companions pursued that through conversation; personal software extends it across the day through many focused applications.
Replika users described ChatGPT, Gemini, and Claude mainly as search, writing, or homework tools, despite rapidly expanding model capabilities. Kuyda concluded that this was an interface problem: when users see a chatbot or command line, those narrow behaviors are precisely what its affordances suggest.
Her metaphor is the “MS-DOS era for AI interfaces.” Almost a billion users make chatbots look widely adopted, but their limited use cases leave room for a Windows-or-macOS-style moment: interactive, visual software through which ordinary people discover what models can actually do.
The macro analogy runs from Russia’s six television channels to YouTube, Reels, and TikTok. Kuyda expects software to follow content from scarce professional production toward UGC—apps built, modified, or suggested by AI around the individual, including a temporary New York art-show finder near the user’s Airbnb.

2. Disposable apps make extreme personalization economic

Long-tail mini apps need not justify businesses or “20,000 features.” One early user built a motivational app that surfaces quotes from a single favorite show at 5:30 a.m.—too specific for an App Store developer, but exactly right for its creator.
At bedtime, Kuyda made her daughter a picture-puzzle game featuring Princess Elsa and Princess Jasmine, then switched it into Italian for preschool practice. “It took me 2 minutes to build it,” followed by seconds of tweaking—versus searching, enduring 15-minute onboarding, paying, and still missing the requested experience.
Justine had deleted more than a dozen apps, some paid for, after replacing them with tailored Wabi versions, from migraine and restaurant tracking to image transformation. Kuyda described similar product-market fit in her lifting tracker: each gym visit adds another revision, so using, creating, and republishing become one continuous behavior.

3. The consumer platform must hide code and contain risk

Anish’s supply-side framing is that only about 20 million developers determine the software everyone consumes. Wabi is therefore designed for nontechnical consumers—not as “text to app” for developers or developer-adjacent users.
When a host suggested Sora-like creation rates might make most users builders, Kuyda disagreed: fully original creators will “probably still” remain under 10%. She expects broader participation through tweaking, remixing, and comments asking creators to modify an app.
At recording time, Wabi was preparing social features showing who was downloading which mini apps, how they were using them, and comments. Its interface deliberately avoids code, API keys, and technical integration language; Kuyda wants Canva-like visual controls and “vibe kind of taste” or “vibe designing,” where users choose a style and concentrate on the use case.
The mobile-first platform also addresses operational risk. Kuyda cited a vibe-coded app around women dating whose sensitive information leaked—not from malicious intent, but because its makers were not professional developers. Like Shopify or a social platform, Wabi aims to supply guardrails, integrations, distribution, and an organizational layer instead of trusting random links and databases.

4. AI-native software compounds context and packages prompts

Early iOS apps merely squeezed websites onto phones or offered toys such as iBeer and the $999.99 I Am Rich. Uber and Tinder arrived after developers exploited mobile-native capabilities; Kuyda believes AI’s equivalent is “really deep personalization,” consistent with Andrej’s “Software 3.0” framing.
Her workout app incorporates a specific training book, a photo of her gym, and the gym’s equipment layout. Separately, at the Wabi platform level, mini apps can know context such as her age, San Francisco location, children, and fitness goals. A future nutrition app should inherit relevant context, replacing today’s “absolutely crazy” walled gardens that require every developer and user to reconnect email, calendar, and other services separately.
Justine’s demand signal comes from teen and college-age girls sharing Nano Banana and Qwen image-edit prompts for scenes such as lying on a couch with the Ghostface killers behind them. The result spreads on TikTok or Reels, while an unwieldy prompt sits in comments and users struggle to determine whether they need the Google or Gemini app.
Kuyda calls it absurd that “godlike technology” is accessed through paragraphs of unstructured commands “like MS-DOS commands—but worse.” A mini app can bundle the prompt, model, reference image, examples, styles, and interaction into a familiar GUI; even a useful blood-work prompt can remain discoverable in a health folder.

5. Apps become creator media and community objects

The investing team’s thesis is that the world has only 1% of the software it needs and the remainder could be built within five years. Kuyda’s extension is to treat apps as content: a fitness creator might publish five mini apps expressing a protocol, perhaps monetize them, and let followers practice together instead of selling another course.
She sees software as the creator economy’s “last frontier.” She points to MrBeast’s close relationship with fans and his chocolate business as an example of the kind of offering creators use to monetize that relationship; mini apps could offer a more interactive connection. Designers could likewise publish aesthetically distinct versions of identical functionality, while Wabi could show what people actually did with a prompt or app beyond impression counts whose viewers may include bots.
Multiplayer could turn apps into communities: parents around local toddler activities, London bird-watchers, or dog owners contributing AI-generated royal portraits to a shared feed. Kuyda wants the result to preserve the early internet’s weirdness—small, raw ideas that could never support App Store businesses—rather than only producing polished, commercial software.

6. Replika proved foresight without scale can still lose

Kuyda’s language-model conviction began in 2012, when a DeepMind friend explained Word2Vec as a way for computers to operate on language. Combining that with Wittgenstein’s “the limits of my language are the limits of my world,” she concluded that learning language meant learning about the world; the team committed to dialogue generation after Google’s August 2015 deep-learning paper by Quoc Le.
They thought the breakthrough was “right around the corner”; it was seven years away. The Meena paper later signaled transformer progress, and in 2020 Mira and Sam showed them GPT-3—a zero-shot, few-shot system that could translate, imitate a tweet, or converse without a separately trained model for each task.
Replika became an early GPT-3 API partner and, Kuyda recalls, the largest customer by API calls; Greg Brockman even trained what she thinks was a fine-tuned Davinci model for it. That position existed partly because Microsoft Tay’s failure had made larger companies afraid to release generative chatbots.
The hosts noted that OpenAI’s later video-game and reinforcement-learning direction was subsequently described by Andrej as a mistaken research path, while Replika stayed with language. Their exchange underscored that being right is not always enough: Anish added that founders must also execute. After raising only $1 million, the company chose scrappy revenue maximization over a contemplated $20 million model bet; Kuyda does not claim the larger bet would have succeeded and says they may not have been the right people, but believes the missed “almost generational chance” carries a lasting lesson.

7. Human empathy points toward a visual AI operating system

Asked how she predicts consumer behavior, Kuyda rejected the premise that she has a special forecasting sense: she has only a few ideas, believes them deeply, and follows them down the rabbit hole. Journalism—starting at age 12 and later as an investigative reporter—gave her practice talking to people and trying to understand their lives.
Watching her computer-savvy mother fail to copy and use a Reddit prompt revealed Wabi’s interface opportunity. Replika came from a parallel observation: lonely people needed someone to hear them, and although AI could not yet talk well, “it could listen”—potentially groundbreaking for millions.
Hardware builders, in her view, fall into a “huge mind trap” by treating voice as the ultimate interface and misreading Her; voice worked there because Scarlett Johansson was “constantly breathing heavily in his ear.” Real users cannot comfortably speak beside a sleeping partner, in an office, in crowds, continually while walking, or in many other ordinary settings.
Even the canonical cooking timer needs a visible countdown; Kuyda says roughly 75% of Alexa devices ship with screens. Voice also handles discovery, proactivity, and notifications poorly because spoken information is slow. Her alternative is a screen-first smartphone with local models, no fixed-app paradigm, and software created on demand: “AI is just an app on your phone. It should not be that way.”

More Truthful AIs Report Conscious Experience: New Mechanistic Research w- Cameron Berg @ AE Studio 余秀华: The Only Thing I Ever Fought Was Marriage, Not Fate