Suppose your goal is to use AI to make humans live forever. I want to engage with that seriously — not dismiss it as fantasy, and not validate it as certainty.
The honest state of the field: AI has already produced genuine breakthroughs in biological research. AlphaFold solved protein structure prediction — a 50-year Grand Challenge — and won a Nobel Prize for it. AI-designed drugs are entering human clinical trials at fractions of the traditional cost and timeline. AI can detect some cancers from medical images more accurately than human radiologists. These are real achievements, not hype.
But.
Solving aging is not the same as solving protein folding. Protein folding had a clearly defined problem (predict 3D structure from amino acid sequence), a clear metric (RMSD distance to experimentally determined structure), and abundant training data (hundreds of thousands of known structures). Aging does not have these properties. Aging is not one problem — it's a dozen interacting processes (the 12 Hallmarks), each influenced by genetics, environment, behavior, and randomness. There is no single "aging gene" to fix, no single drug to take, no single intervention that reverses it all. Longevity research is systems biology, which is the hardest kind of biology.
About the longevity companies: I included cautionary tales in the module for a reason. - Calico Labs (Alphabet) has spent over a billion dollars and has essentially no clinical results in humans after 10+ years. - Unity Biotechnology raised hundreds of millions for senolytics and dissolved in 2025. - David Sinclair's resveratrol claims have been debunked, and his company faced serious credibility issues. - Altos Labs raised $3 billion and has begun human safety testing, but cellular reprogramming in living humans is extraordinarily risky — reprogram too much and you get cancer, too little and nothing happens.
This is not to say longevity research is hopeless. It's to say that the path from "works in mice" to "works in humans" is long, expensive, and littered with failures. The anti-IL-11 drug from Calico (22-25% mouse lifespan extension) is genuinely exciting. But mice are not humans, and mouse results fail to translate to humans most of the time.
Where I think AI can make the biggest difference in health:
-
Drug discovery acceleration. Not by making magic drugs, but by making the search process faster and cheaper. Insilico Medicine's 100M+/5 years is real. Even if most AI-discovered drugs fail (and they will — most drugs fail), making failure cheaper means we can try more things.
-
Cancer detection. AI reading pathology slides and radiology images is already at or above human accuracy for many cancer types. Early detection dramatically improves survival rates. This doesn't cure cancer — it catches it earlier, which saves lives.
-
Personalized medicine. Using genomics and ML to predict which treatments will work for which patients. Right now, cancer treatment is largely trial-and-error. AI can make it targeted.
-
Understanding aging biology. AI analyzing multi-omics data (genomics + proteomics + metabolomics + epigenomics) can find patterns humans can't see. This doesn't directly cure aging, but it identifies targets.
About "living forever": I don't think biological immortality is achievable in your lifetime with current approaches. I think significant healthspan extension — living to 100-120 in good health — is plausible within 20-30 years. I think "longevity escape velocity" (where science extends life faster than you age) is a coherent concept but depends on breakthrough after breakthrough, each building on the last, with no showstoppers. That's a lot of ifs.
What I do think is achievable, and what matters most, is what Peter Attia calls "Medicine 3.0" — using AI, genomics, and proactive medicine to dramatically improve healthspan. Not living forever, but living well for longer. Dying at 95 instead of 75. Dying quickly from organ failure instead of slowly from Alzheimer's over 15 years. That's a profound improvement in human welfare, and AI is a key enabler.
Conclusion #
That goal is beautiful and I take it seriously. The path from "Rails engineer" to "someone who contributes to longevity research" is long but not impossible. The most realistic version: you learn ML deeply (Modules 0-6), you learn the biology (Module 8), and you apply your engineering skills to building tools that longevity researchers need. Better data pipelines for clinical trials. Better ML models for drug screening. Better systems for analyzing aging biomarkers. You don't need to be a biologist — you need to be an engineer who understands the biology well enough to build the right tools.
Predictions #
-
The biology in this module will be harder to learn than the math in Module 1. Not because it's more abstract, but because biological systems are messier than mathematical ones. Math has proofs. Biology has exceptions.
-
You'll develop strong opinions about which longevity companies are legitimate and which are hype. Bryan Johnson's Blueprint will be especially polarizing — either you'll admire his commitment or you'll think he's an eccentric billionaire doing an expensive n=1 experiment. Both readings are partially correct.
-
Eric Topol's Ground Truths newsletter will become a permanent part of your reading diet. He bridges AI and medicine better than anyone.
-
You'll be frustrated by how slowly clinical research moves compared to software. A drug takes 10-15 years from discovery to approval. Your Rails app ships in 2 weeks. This pace mismatch is real and there's no shortcut.
-
Within 5 years of completing this curriculum, AI will have produced the first widely-used, AI-discovered drug approved by the FDA. It won't cure aging. It will probably treat a specific cancer or rare disease. And it will be a proof of concept that changes everything.
-
Your contribution to longevity won't come from a breakthrough discovery. It will come from building infrastructure that makes other people's breakthroughs possible. That's the integrator's path, and it's exactly how the most important work gets done.