Episode Summary
Executive Summary: Jacob Kimmel, president and co-founder of New Limit, argues that evolution under-selected for longevity and late-life intelligence because high mortality, kin selection tradeoffs, and optimization constraints made aging hard to solve. He outlines New Limit’s strategy: use epigenetic reprogramming and AI-guided transcription-factor discovery to restore youthful cell states, while solving delivery and reimbursement challenges needed to turn these insights into therapies.
Main Topics: Why evolution did not optimize for longevity (Priority: 5/5): Kimmel argues that high baseline mortality during human evolution reduced the selective pressure for long lifespans, while tradeoffs and optimization constraints further discouraged evolution from 'solving' aging. Age, intelligence, and developmental timing (Priority: 4/5): The discussion connects longevity to intelligence, suggesting fluid intelligence peaks at ages where evolutionary selection was strongest, helping explain why major discoveries and achievements often happen relatively early in life. Epigenetic reprogramming as a rejuvenation strategy (Priority: 5/5): New Limit’s core thesis is that aging is partly an epigenetic problem: cells retain the same DNA but lose youthful regulatory programs, which can potentially be restored by transcription factors. AI/ML for cell-state modeling and perturbation prediction (Priority: 5/5): The interview frames biology as a combinatorial search problem: models can learn how transcription-factor perturbations shift cell states, enabling a 'virtual cell' approach to discovery. Delivery as the bottleneck for future medicines (Priority: 5/5): Kimmel argues the major challenge is not just discovering the right factors but delivering them to the right cells. He contrasts lipid nanoparticles, viral vectors, and future engineered cell-based delivery systems. Drug discovery, Eroom’s Law, and pharma economics (Priority: 4/5): The conversation turns to why biotech productivity has worsened over time, how AI-style general platforms could improve returns, and how reimbursement may shift toward pay-for-performance or direct-to-consumer models.
Key Arguments: Evolution likely did not strongly select for longevity because the hazard rate in ancestral environments was extremely high, so few individuals lived long enough for late-life traits to be strongly rewarded. Longevity may be selected against in part because aged individuals can become net negative for genome propagation if they consume resources without fully preserving fitness. Aging is probably not monocausal; it reflects layered molecular regulation, especially epigenetic state changes, so the likely therapeutic path is multiple interventions rather than a single magic reset. Transcription factors are attractive drug targets because they are the genome’s regulatory levers, but direct targeting has historically been hard due to delivery and modality limitations. Single-cell genomics makes it possible to measure whether a cell is becoming younger while also checking for unwanted changes like loss of identity, hyperinflammation, or neoplasia. The Yamanaka factor discovery worked because the target state was easy to identify and the successful phenotype amplified; aging is harder because success is subtler and does not self-amplify. AI-style models can turn sparse perturbation data into predictive maps of cell-state change, similar to how foundation models learn general representations before task-specific optimization. The long-term future of drug delivery may resemble the immune system: engineered cells that sense context and release payloads only when needed. Even partial rejuvenation of key tissues may produce broad systemic benefits because organs like liver and immune cells are highly interconnected with whole-body physiology. Pharma’s future may be platform-based, but commercialization will likely still depend on IP, payer structures, and the ability to prove long-term clinical benefit.
Data Points: Baseline hazard rate during evolution: Very high - Used to explain why few ancestral humans reached old age, limiting selection for longevity Age of peak fluid intelligence / discovery: Roughly 25-30 - Kimmel’s hypothesis for when selection for preserved fluid intelligence may have been strongest Great scientific discoveries: Often before 30 - Cited as supporting evidence that achievement peaks relatively early in life Yamanaka factors: 4 transcription factors - Classic iPSC reprogramming set that converts adult cells back toward stem-cell state Human TF count: ~1,000 to 2,000 transcription factors - Used to estimate combinatorial search space for reprogramming Possible TF combinations for 1-6 factors: About 10^16 - Illustrates why brute-force screening is infeasible iPSC reprogramming efficiency: ~0.01% to 0.001% - Kimmel cites original Yamanaka work as a low-efficiency but amplifiable discovery Single-cell sequencing cost: From dollars to cents/fractions of cents - Technological improvement enabling much larger perturbation studies Perturb-seq detection accuracy early on: ~50% barcode detection - Early technical limitation in assigning perturbations to cells Cellular perturbation scale: Millions of cells per day - Current throughput described for New Limit experiments Medicare spending concentration: ~One-third in the final year of life - Used to argue that preventing late-stage decline could lower overall healthcare costs U.S. healthcare spend on drugs: ~7% - Cited as an approximate share of total healthcare spending Biotech output metric: Eroom’s Law - Described as a long-run decline in new molecular entities per billion dollars invested HRT market share: About 15% of U.S. equities volume - Mentioned in a sponsor segment about AI/trading infrastructure, not part of the scientific discussion
Pivotal Quotes: "I think actually the lifespan argument plays back into intelligence to a degree." — Jacob Kimmel: Explaining how longevity and intelligence may be evolutionarily linked through selection timing "The epigenome can degrade with age. It changes." — Jacob Kimmel: Summarizing the biological basis for New Limit’s rejuvenation strategy "We actually need to search a much broader portion of TF space in order to be successful." — Jacob Kimmel: Why aging reprogramming requires AI and large-scale perturbation data rather than a small manual search
Implications: The conversation suggests aging may be treatable through targeted, multi-factor cell-state reprogramming, but progress depends on AI-guided discovery, precise delivery, and proof of durable clinical benefit. If successful, medicine could shift from disease treatment to proactive healthspan extension.