A Guy Used AI to Cure His Dog's Cancer*

What It Covers

Nathaniel Whittemore examines why AI discourse has reached peak intensity in early 2026, using two viral weekend stories — Andrej Karpathy's job displacement visualization and an Australian man's AI-assisted cancer vaccine for his dog — to illustrate the widening gap between mainstream AI perception and actual capability.

Key Questions Answered

• •AI's Second Moment Framework: Recognize that 2025-2026 represents a distinct second inflection point in generative AI, comparable to the ChatGPT moment of late 2022. The first moment centered on language models; this moment centers on agents. Understanding this framing helps contextualize why discourse intensity has escalated disproportionately across media, politics, and financial markets simultaneously.
• •Exposure ≠ Displacement: Karpathy's job visualization scored 342 occupations on AI exposure zero-to-10, with software developers at 8-9 and medical transcriptionists at 10. Karpathy himself clarified the tool measures digital exposure, not replacement probability. Economists note high-exposure jobs historically attract increased hiring and wages due to demand elasticity — a critical distinction to apply when evaluating AI labor market predictions.
• •SEC Filings as Leading Indicator: Track SEC 10-K filings rather than CEO statements for honest AI risk assessment. In early 2025, 27 firms listed AI agents as material business risks, up from 7 the prior year. Companies including Figma, Workday, and HubSpot filed agent-risk disclosures while their own CEOs publicly downplayed the same threat in earnings calls.
• •AI as Corporate Cover: Investor Chamath Palihapitiya's framework suggests AI is functioning as plausible deniability for companies executing workforce reductions they already planned post-COVID over-hiring. When evaluating company layoff announcements citing AI, consider whether the technology is the actual driver or a convenient justification for corrections that were financially necessary regardless.
• •Personalized mRNA Vaccine Pipeline: Australian entrepreneur Paul Koyningham used ChatGPT prompting plus Google DeepMind's AlphaFold to identify tumor protein mutations in his dog Rosie, enabling University of New South Wales researchers to design a bespoke mRNA cancer vaccine in under two months for roughly $3,000 in sequencing costs. The lead researcher argues this demonstrates a replicable democratized model for human personalized cancer treatment.