Sarah Guo

Sarah Guo × Max Hodak

Prima won CE marketing approval in Europe in July, with commercial sales expected in coming weeks after trials let patients with macular degeneration complete Sudoku, crosswords, and read books.Science acquired Pixium’s Stanford-originated implant after about two years, calling it “by far the state-of-the-art”; pricing is unset, while future versions could improve color and expand the population from hundreds of thousands to millions.

Elad Gil × Sarah Guo × Mark Zuckerberg × Priscilla Chan × Alex Rives

Biohub’s $500 million, 10- to 15-year commitment targets purpose-built biological data, combining frontier AI with wet labs from proteins to cells and whole systems.ESMfold predicted structures for more than 1.1 billion proteins and produced nanomolar binders from 96 synthesized designs without antibody-specific training, while open release could accelerate research but leaves biosafety and clinical translation unresolved.

Sarah Guo × Elad Gil

AI adoption is “blinding” among doctors, lawyers, and compliance teams, even as a weak Nvidia quarter could trigger panic; verticals such as coding, medical scribing, and legal software such as Harvey are consolidating.Monitor robotics timeline slippage and whether an AI IPO becomes a reflexive benchmark trade; Sarah forecasts somebody will make “hundreds of millions of dollars trading markets with LLMs” in 2026.

Sarah Guo × Sajith Wickramasekara

Biotech is shifting from platform hype toward speed, cost and better molecules, as Chinese companies produce clinical-stage assets “really fast and really cheap.”Benchling’s structured records and agents can recover lost institutional knowledge, recommend experiments and compress workflows, while predictive models remain constrained by trust, IP and regulation.The key catalyst is cumulative pipeline compression—from “7 to 10 years down to 2 to 3 years”—rather than a single AI-discovered drug.

Daniel Nadler × Sarah Guo × Elad Gil

OpenEvidence says it became the default clinical-knowledge system for U.S. physicians within 18 months, reaching roughly 40% of doctors and about 20 times the usage of the next clinical decision-support platform.Its free, physician-focused distribution and semantic search across 35 million biomedical publications turn evidence retrieval into a scalable “curbside consult,” while exploding medical knowledge makes continuous decision support increasingly structural.

Sarah Guo × Joshua Meier × Jack Dent

Chai-2 produced antibody hits in roughly 20% of up to 20 designs, versus about 0.1% or lower for earlier computational methods, with at least one hit on roughly half of more than 50 held-out targets.The broad benchmark and roughly two-week lab cycle suggest AI can expand addressable discovery, but the results remain hits rather than drugs: manufacturing, stability, capital markets and clinical risk are still unresolved.

Sarah Guo × Elad Gil

AI leaders are emerging in foundation models, coding, medical scribing, and customer success, while sales productivity, financial-analyst tooling, accounting, and pharma remain open.The winning application playbook combines vertical workflows, proprietary data, distribution, and user-generated knowledge, while startup mergers may strengthen defense against incumbents; neglected fertility, longevity, and world-model opportunities remain constrained by financing, regulation, and uncertain commercialization.

Sarah Guo × Noubar Afeyan

Noubar Afeyan’s Flagship Pioneering model treats parallel startup creation as a repeatable institution for underwriting uncertainty, using variation, selection and iteration rather than forecasting breakthroughs.Generate Biomedicines has more than 15 computationally designed antibody programs, but Phase III trials, FDA regulation, pricing power and platform funding remain decisive constraints, as Joule’s technically successful but economically unviable outcome demonstrates.

Sarah Guo × Nima Alidoust × Johnny Yu × Patrick Hsu × Dave Burke × Hani Goodarzi

Tahoe 100 brings 100 million single-cell measurements across 50 cancer models, 1,200 drug treatments, and 60,000 interactions to machine-learning scale.Its causal, diverse, consistently generated data supports Vivo’s Mosaic platform, while open sourcing recruits external model development.Virtual cells could narrow target and chemical search, but current differential-expression prediction is around 10%, benchmarks are unsettled, and therapeutic proof may take a roughly 10-year window.