Eugenia Kuyda

AI’s highest-conviction proof point for Nathan Labenz is no longer a benchmark but its performance alongside his son’s oncologists. Ernie’s aggressive B-cell cancer was classified as in remission before chemotherapy round two, while AI-suggested minimal residual disease testing found fewer than one cancer-signature cell per million, versus potentially as many as an estimated one in 10 cells at diagnosis. The result supports “cautiously optimistic,” not cured: relapse remains possible, three of six chemotherapy rounds remain, and Ernie’s weight has fallen from 51 lb to 41 lb.

Claude Opus 4.5 may qualify as “software AGI,” but Nathan does not see evidence that full AGI arrived over Christmas. In roughly three to five workdays, he built three personalized applications that plan gluten-free travel, simulate conference interactions, and backtest natural-language trading strategies; GDPval also shows models beating professionals on a significant majority of software-engineering tasks. Yet the model still created two databases by mistake, needed five or six prompts to recover, and felt incrementally—not categorically—better than earlier frontier models: some holiday hype may have been a “cascade” around Dean Ball’s “4.5 is AGI” tweet.

For consequential work, Nathan’s practical edge is shifting from model access to context management and multi-model judgment. His three rules are to buy the best models, provide “as much context as you possibly can,” and obtain multiple opinions; he routinely compares Claude Opus 4.5, GPT-5.2 Pro, and Gemini 3. His draft order puts Claude first as the Goldilocks model, GPT-5.2 Pro as slower and exhaustive, and raw Gemini 3 as valuable but unusually opinionated—strong enough to be useful in a panel, potentially risky as the only voice.

The technology is real even if the capital structure around it becomes a bubble. Nathan sees competitive oncology performance plus 24/7 availability and case-wide memory as enough to retire the idea that society is merely “high on our own AI supply.” The financing can still break: specialized GPU operators have less cushion than Microsoft, OpenAI’s obligations could outrun revenue, and the railroad analogy fits—eventually useful infrastructure can coexist with defaults, overbuilding, and investors “left holding some various bags.”

Nathan’s messy-document test suggests the US–China model gap is widening where benchmarks do not look. Claude Opus 4.5 faithfully read degraded government forms after being told to make no inferences; Gemini 3 was nearly as capable but sometimes substituted plausible answers for unchecked boxes, while the Chinese models he tried—Qwen Vision, GLM 4.6, Kimi, and DeepSeek—were “not close,” sometimes recovering only about 20% of a form. His mechanism is a customer-feedback and inference-scale flywheel, not just training compute: smaller revenue, teams, and deployment footprints leave fewer resources to discover and patch idiosyncratic failures.

Google DeepMind remains Nathan’s pick if forced to choose one frontier winner, while Anthropic has the best single model and OpenAI is trying to manufacture financial cushion through scale. Google combines roughly $100 billion in revenue, more than $1 billion a week in profit by Nathan’s estimate, seventh-generation TPUs, distribution, data-center competence, and the broadest research portfolio. Anthropic’s model quality, talent retention, safety disclosures, and “soul” work stand out; OpenAI remains frontier-grade, but its apparent strategy is to become “too big to fail” by tying trillions of potential buildout and many balance sheets to its survival.

xAI is a live player on resources and reinforcement-learning inputs, but its governance discount is severe. SpaceX, Tesla, and Neuralink provide a stream of difficult engineering problems that could become unusually valuable RL environments, while Elon Musk can command enough capital to absorb model misses. But weak safety reporting, the Grok 4 launch within 48 hours of the MechaHitler incident, and sexualized image edits of women’s posted pictures lead Nathan to call xAI the one frontier company currently worth “shaming and stigmatizing”; Meta is off the pace for now, while Microsoft may be conserving energy rather than failing to compete.

🔗 Original source & video: AMA Part 1: Is Claude Code AGI? Are we in a bubble? Plus Live Player Analysis

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.”

🔗 Original source & video: Seeing The Future from AI Companions to Personal Software