Ramaneumann
3.7K posts



When you do any kind of science, when you get to the edge, you find things that are counterintuitive. If they had been intuitive, they would have been discovered long ago. Science is about understanding things that go beyond our intuition. ~Conjecture Institute Advisor @DavidDeutschOxf with @GadSaad


One week of short sleep in otherwise healthy adults. Not "sleep deprived" by strict definition. Just 4-5 hours a night. Testosterone dropped 15%. Insulin sensitivity dropped 20%. Muscle protein synthesis dropped 19%. Hunger hormones rose 28%. Cortisol rose 51%. These aren't the only systems affected. They're just some of the ones that have been measured in controlled settings. No supplement, no diet hack, no training program (crazy claim, I know, but you can't outrain poor sleep...) outperforms sleep at keeping systems "online". References: Leproult & Van Cauter, JAMA, 2011. Buxton et al., Diabetes, 2010. Zuraikat et al., Diabetes Care, 2024. Spiegel et al., Lancet, 1999. Saner et al., J Physiol, 2020. Spiegel et al., Ann Intern Med, 2004.

In today's newsletter @jackclarkSF predicted that full no-human-involved AI R&D will happen by the end of 2028. Much of the pushback against RSI has been that AI has not yet shown the capacity to generate fully new ideas. This is the key part from Jack's post: the majority of AI research doesn't need for this to happen---RSI in AI research can just be driven by `meat and potatoes' engineering work. open.substack.com/pub/importai/p…


Anthropic is about to announce a $1.5 billion joint venture with multiple wall street firms to sell AI tools to private-equity backed companies. Anthropic, Blackstone, Hellman & Friedman and Goldman Sachs are all major investors. The announcement is expected tomorrow morning.

Meanwhile Mission and 23rd in San Francisco. People have learned that placing random objects into the wheel well could disable the autonomous vehicle.



This is free btw. Fantastic listen.



A common feminist misinfo misrepresents recent paternal age birth risk research. Comparing ages 20s vs 50s, Dads have tiny absolute risk increases (~1%-3%) for non-fatal risks like ADHD/autism, but Moms have largest absolute risk increases for fatal risks like miscarriage (~11%-50%+). Evolution (having offspring) tolerates the tiny rise of nonfatal birth risks in older fathers but not the large rise of fatal birth risks in older mothers. Which is reflected in traditional lindy gender/age pairing differences. Bookmark this for reference next time you see paternal age misinformation! (I had Claude's top model summarize the research) claude.ai/share/bb1464f1…





New composite eval of DeepSeek V4 from CAISI suggests China is falling behind. Notice the relative steepness of their improvement trend. nist.gov/news-events/ne…

David Reich on how much ancient DNA evidence has overturned so much consensus thinking how ancient cultures spread. "It wasn't peaceful, it wasn't friendly, it wasn't nice. Some of our archaeologist co-authors were just really distressed."




A recent viral paper claims to reverse-engineer the parameter counts of frontier models: GPT-5.5 = 9.7T, Opus 4.7 = 4.0T, o1 = 3.5T, etc. @ben_sturgeon and I investigated and found serious issues in the paper; fixing them gives GPT-5.5 as ~1.5T (90% CI: 256B-8.3T).



@opencode + @deepseek_ai v4 pro = solid combo











