Dwarkesh Podcast

David Reich – Bronze Age shock, the Neanderthal puzzle, & the sudden spread of farming

David Reich is back. He and collaborator Ali Akbari just published a paper that overturns a long-standing consensus about human evolution — that natural selection has been dormant in our species since the agricultural revolution. By scaling ancient DNA sequencing and developing a new statistical method, they found that selection has actually sped up. Selection went especially bonkers during the Bronze Age (around 3,000 years ago). That's when gene frequencies for everything from immune function to body fat to intelligence were most in flux. Over the last 10,000 years, selection pushed the genetic predictor of cognitive performance up by roughly a full standard deviation — most of it between 4,000 and 2,000 years ago. After we finished recording, David sketched out on a whiteboard his new heretical model about who the Neanderthals really were. Luckily, I took out my iPhone and managed to record it. He thinks the standard story (that Neanderthals are some separate archaic lineage we interbred with a little) just doesn't fit the evidence. Instead, he proposes that Neanderthals are essentially genetically-swamped modern humans. A small population somewhere around the Caucasus invented Middle Stone Age technology roughly 300,000 years ago and expanded outward. The ones that moved into Europe interbred with local archaic humans, got genetically swamped, and became Neanderthals. The same expansion went into Africa, met much more diverged archaic Africans, and that mixture became us. This means Neanderthals and modern humans share the same cultural ancestry — the only difference is which archaic humans they mixed with afterward. David is a brilliant and rigorous scholar. It was a real delight to learn from him again. +𝐄𝐏𝐈𝐒𝐎𝐃𝐄 𝐋𝐈𝐍𝐊𝐒 * Transcript: https://www.dwarkesh.com/p/david-reich-2 * Apple Podcasts: https://podcasts.apple.com/us/podcast/david-reich-why-the-bronze-age-was-an-inflection/id1516093381?i=1000766816517 * Spotify: https://open.spotify.com/episode/6BZ56Puv0gsnWCA8yfSde4?si=s7fv1yuuR5ykDMyEIcPTeQ 𝐒𝐏𝐎𝐍𝐒𝐎𝐑𝐒 - Cursor was super useful as I prepped for this episode. Whenever I had a question, I'd have Cursor kick off a few different models simultaneously and then compare their responses. I found that this led to better results than I could get out of any individual LLM. If you've only used Cursor for coding, you should try using it for research. Check it out at https://cursor.com/dwarkesh - Jane Street uses an internal currency called "hive bucks" to allocate compute through a real-time auction – and anyone can change anyone else's bids or even kill their jobs! Everyone just trusts each other to act in the firm's best interest, which is what lets the system work in the first place. If this weird and high-trust culture sounds like your kind of thing, Jane Street's hiring at https://janestreet.com/dwarkesh - Crusoe's ML infra team built fastokens, an open-source tokenizer that delivers a ~9x speedup over Hugging Face and up to 40% faster time-to-first token – on real production workloads! Crusoe achieved these results by parallelizing things and using some clever engineering to handle duplicates without cross-thread coordination. Learn more at https://crusoe.ai/dwarkesh To sponsor a future episode, visit https://dwarkesh.com/advertise. 𝐓𝐈𝐌𝐄𝐒𝐓𝐀𝐌𝐏𝐒 00:00:00 – Ancient DNA suggests strong selection over last 10,000 years 00:16:24 – Natural selection intensified during the Bronze Age 00:35:40 – Why didn't evolution max out intelligence? 00:58:00 – Evolution is limited by time, not population size 01:09:40 – Why no farming before the Ice Age? 01:17:52 – The Neanderthal puzzle David can’t stop thinking about 01:54:40 – The methodology behind this breakthrough

DRDavid ReichguestDPDwarkesh Patelhost

May 8, 2026·2h 13m·Watch on YouTube ↗

📖 CHAPTERS

1. 1

   Ancient DNA finally gets big enough to measure selection over time

   Reich explains why ancient DNA revolutionized human migration history long before it could deliver on the original promise: tracking biological change. The key bottleneck was sample size—single genomes are rich for ancestry, but poor for estimating allele-frequency trajectories and selection.

2. 2

   What allele-frequency change tells you (and why migration is a confounder)

   They discuss why frequency shifts across time are informative about adaptation to changing environments (diet, altitude, pathogens, domestication). Reich emphasizes that most frequency change comes from migration/admixture rather than selection, so selection signals must be detected as locus-specific deviations from genome-wide shifts.

3. 3

   A surprising result: selection is widespread but usually subtle

   Reich summarizes the headline finding: the genome is ‘vibrating’ with selection even if selection explains only a small fraction of total allele-frequency movement. They estimate thousands of candidate selected positions, with varying confidence thresholds, revealing selection is far from quiescent in the last 10k years.

4. 4

   What traits are most targeted: immunity and metabolism dominate; behavior is harder

   By intersecting selection signals with GWAS traits, the strongest enrichment appears for immune-related variants and metabolic traits (obesity/diabetes-related). Behavioral and psychiatric traits show less enrichment in the ‘top hits’ not because they aren’t selected, but because they are extremely polygenic with tiny per-variant effects.

5. 5

   The Bronze Age inflection: accelerated selection in the last ~5,000 years

   A major theme is that selection intensifies in the Bronze Age and afterward, more than during the initial adoption of farming. Reich frames this as a ‘shock’ from rising population density, new pathogens, animal proximity, and intensified social/technological changes, creating evolutionary mismatch and rapid adaptation.

6. 6

   Concrete examples of shifting selection pressures (TB, FADS, ABO, iron, pigmentation)

   They walk through specific loci showing striking time dynamics, including reversals where an allele rises then falls. Examples include TB risk (TIC2), dietary fat metabolism (FADS1/2), ABO blood groups, hemochromatosis-related variants, and depigmentation timing peaking around 4,000–2,000 years ago.

7. 7

   Polygenic selection on cognition/education signals—and why interpretation is tricky

   Reich claims strong polygenic selection on predictors of cognitive performance and years-of-schooling, peaking in the Bronze Age and largely absent in the last 2,000 years. They stress the measured predictors may proxy broader traits (executive function, planning, fertility timing) rather than ‘IQ’ per se, and validate robustness using cross-population GWAS comparisons.

8. 8

   Why evolution didn’t ‘max out’ intelligence: tradeoffs, changing optima, and fertility dynamics

   They explore why seemingly universally useful traits might not monotonically increase. Reich speculates that selection may act on multidimensional tradeoffs (e.g., quality vs quantity of offspring investment), and that some psychiatric risk alleles could be linked to advantageous subclinical traits in certain cultural contexts.

9. 9

   Selection on body fat and metabolism: the ‘thrifty genes’ debate and timescales

   They discuss evidence for selection against obesity/BMI-associated variants over the last 10,000 years in Europe/Middle East. Reich connects this to the thrifty genes hypothesis and argues the relevant stability may be short-term food access (boom-bust hunting) rather than multi-year famine dynamics common in agricultural societies.

10. 10

    Time vs population size: why bigger Bronze Age populations aren’t the main explanation

    Dwarkesh asks whether Bronze Age population growth made selection more effective by generating more mutations and overcoming drift. Reich argues strong selection (≈0.5–1%+) works even in small populations; for weak selection that depends on huge population size, the timescales would be far too long to matter here—time is the binding constraint, not N.

11. 11

    Why no farming before the Holocene: climate stability as the missing ingredient

    Despite genetic ‘readiness,’ agriculture appears only after ~12,000 years ago and then emerges independently in multiple regions. Reich highlights a puzzling claim from climate science: the Holocene brought unusual long-run climate stability compared with the prior two million years, potentially enabling sustained cultivation and domestication.

12. 12

    The Neanderthal puzzle: genomes say one thing, archaeology says another

    Reich describes a persistent tension: genome-wide, Denisovans and Neanderthals are sisters, yet Neanderthals share many cultural and genetic features with modern humans. He focuses on anomalies like Neanderthal mitochondrial DNA and Y chromosomes clustering with modern humans, plus shared Middle Paleolithic/Levallois technology absent in East Asia.

13. 13

    A new speculative model: an early modern-human expansion reshaping Neanderthals (and us)

    In a whiteboard-style explanation, Reich proposes an alternative framing: a Middle Stone Age/Levallois innovation spread via an expansion that mixed heavily with local Eurasian archaics, leaving small genome-wide ancestry but potentially replacing uniparental lineages and transmitting culture. He draws an analogy to epicycles—suggesting current models may be over-patched—and notes parallel evidence for deep African substructure and admixture into modern humans.

14. 14

    Methodological breakthrough: predicting genotypes from relatedness, then testing for selection

    Reich returns to the technical core: they model expected allele states using a relatedness matrix across ~22k individuals (ancient + modern), capturing drift and ancestry shifts, then test whether adding a consistent directional-selection term improves prediction. Large data scale plus this statistical framing yields far more power than earlier ‘ancient vs modern frequency difference’ scans.

15. 15

    Validating signals via GWAS enrichment; guarding against background selection

    To calibrate which selection statistics are likely real, they use an external validation: enrichment for GWAS-associated loci rises with selection score and plateaus, implying high scores are mostly true positives. They test alternative explanations like background selection (purifying selection near genes) by stratifying genomic regions to show the enrichment persists.

16. 16

    How ancient DNA scaled: cheaper sequencing + capture enrichment + industrialized pipelines

    Reich explains the practical innovations enabling the new sample sizes: dramatic sequencing cost declines, plus in-solution capture that enriches human DNA from microbe-dominated remains. Roboticized pipelines and targeted panels made it economical to generate thousands of samples per year, transforming what questions the field can answer.