r0ze-at-hn

Coase’s math assumes that the intervals between price changes approach zero, but every state change (such as the ebook price updates) is an informational bit erasure or write event. In a real market there is a cost that is > 0 for any price change (landauer floor). One can not assume that there is "free" energy in the system to find the optimal price.

Anyone writing books (or a "firm in coses math) needs to persist, aka they need to keep things like consumers understanding of their quality and pricing. If they drop prices to nothing they have consumers learn that. Consumers that try to wait forever in an "idle" state make no purchases and are not part of the ecosystem. Only consumer that don't wait forever matter. aka if they are looking at your ebook and don't buy they will rapidly buy a different ebook because they cannot wait for infinite time. aka if you teach your customers to wait they wont actually wait for you, but will simply switch to something else.

This is a control-and-feedback problem. Coase Conjecture fails because it assumes that you can have a system that persists without a governor. The two real world explanations correspond to the two ways you can introduce a governor.

It is also very extensively tested against the official WASM testsuite.

I was hoping that near the end the author would have tried to contributed any new tests to the official testsuite to help catch these same errors elsewhere.

For the first one, Zero Is Not Null maybe there is a missing test in in the call_indirect test?

https://github.com/WebAssembly/testsuite/blob/main/call_indi...

Typically it is expected that a software engineer gets their first promotion between 12-24 months. At the 6-12 month timeframe the managers will be having discussions around if they are on track, what they need help on (everyone needs a little help), or if there clear performance warnings going on and we need to take action of some sort.

I will congratulate everyone on their first promotion, it is worth celebrating, not everyone can do this job. But this first promotion is given to everyone who can actually do the work.

Get someone good, in a greenfield project, the right start timing aligned with promo committee time, add in some luck, and sure two promos in two years are possible.

Seen this before and the worry is that they are learning is the game of the promos system, not engineering. I would have to sit them down and ask them how many years away from being the CEO do they think they are. The next promo will come slower no matter their skill and even slower after that and one day the promos will probably stop. The validation that they might be addicted to will get harder to obtain over time. Not getting a promos shouldn't "crush" you. If they assume that merit is always rewarded they should ask their boss how promo committees actually work. Many people don't get promos for reasons unrelated to merit.

The second question is where is the engineering? Only two years into a long career there is a lifetime to learn and explore. If they only chase promo they are going to burn out very quickly.

Linking to the paper: https://arxiv.org/pdf/2605.01172 which is also a fantastic read, the application to deep learning is good. It does a lot of cross-mapping and highlighting a bunch of old stuff that is named differently in this paper and worth calling out for those with those backgrounds:

"Cumulative Dissipation Gramian" Ws = Observability Gramian (from Control Theory). For example the spectral cutoff is exactly the Hankel singular value truncation from model reduction.

"Signal Channel" / "Reservoir" is Controllable/Observable vs. Uncontrollable/Unobservable Subspaces. Using Adamjan-Arov-Krein (AAK) theory gives the optimal nonlinear reduced model answering the optimal compression question.

"Drift–Diffusion Separation" is Freidlin-Wentzell Large Deviation Theory. They can predict "grokking" time from the FW action.

"Population-Risk Gate" is Quantum Weak Value / Postselection (Aharonov)

So for the follow-up problems

Control theory gives the truncation error bounds for model compression. Large deviation theory gives the grokking time predictions. Quantum measurement theory gives the imaginary preconditioners. Information geometry gives the optimal continuous relaxation of the gate.

Some nice implications of new ways of doing stuff which are nice to see formalized here:

Old: Pick architecture, hope it generalizes New: Design architecture to maximize observability Gramian rank (Honestly we pull a lot from control theory here)

Old: Use validation set to detect overfitting New: Monitor λ(Ws) spectrum during training; no validation needed

Old: Prune post-hoc based on magnitude New: Prune during training based on ker(Ws) membership

Old: Fixed learning rate New: Spectral learning rate

There has been massive progress in neurodegeneration research. We now understand that the brain has a complex system for processing metabolic "trash." Like all systems it has a number of components, each of which can fail. This is why a single "silver bullet" drug will likely never exist. A multi variable systems failure can not be fixed with a single variable solution.

Alzheimer’s occurs when Amyloid/Tau debris accumulates faster than the glymphatic system can export it.

Parkinson’s occurs when α-synuclein (Lewy Bodies) accumulates because the cellular recycling system (Mitophagy) has failed.

While Choline is a critical bottleneck for Alzheimer's, Glutathione and GBA enzyme activity are the primary bottlenecks for Parkinson’s. But in both cases it depends on the person and their genetics and what matters the most. If someone is at high risk for Alzheimer's a multi-pronged approach will probably be very common. For example a post menopause woman who is not on HRT, but at high risk due to parents should be getting Choline due to the low PEMT expression, but they might have a higher risk for neuroinflammation via NLRP3 and so need to also combat that too.

To say no progress has been made is to ignore the fact that we have learned a ton about the various components of these systems and how they can break down.

Here are the two approximate decay equations that you can put into an online latex viewer.

Alzheimer's

\Psi_{AD} = \int_{0}^{t} \frac{[(\mathbf{K}_{\tau} \otimes \mathbf{K}_{A\beta}) \cdot \mathcal{I}_{NLR P 3}]}{[\mathcal{G}(Li \cdot GSK3\beta^{-1}) \cdot PRO(ER\alpha \cdot \text{PEMT}) ] \cdot \Omega_{ATP}} \cdot e^{\left( \frac{\Gamma_{\text{fibrosis}} \cdot \text{PAI-1}}{EPA \cdot \text{Natto}} \right)} \, dt

Parkinson's

\Psi_{PD} = \int_{0}^{t} \frac{[(\mathbf{K}_{\alpha\text{-syn}} \otimes \mathcal{I}_{LRRK2}) \cdot \mathcal{I}_{NLRP3}]}{[\mathcal{G}(\text{Parkin} \cdot \text{PINK1}) \cdot PRO(\text{Dopamine}_{\text{flux}})] \cdot \Omega_{ATP}} \cdot e^{\left(\frac{\Gamma_{\text{Oxidative}} \cdot \text{Iron}}{\text{Glutathione} \cdot \text{GBA}}\right)} \, dt

And for good measure here is cancer

\Psi_{Onco} = \int_{0}^{t} \frac{[ (\mathbf{M}_{mut} \otimes \mathcal{G}_{growth}) \cdot \mathcal{I}_{STAT3} ]}{ [ \mathcal{G}(p53 \cdot \text{PTEN}) \cdot PRO(\text{Anoikis}) ] \cdot \Omega_{ATP}} \cdot e^{\left( \frac{\Gamma_{hypoxia} \cdot \text{VEGF}}{\text{Repair}_{DNA} \cdot \text{Immune}_{flux}} \right)} \, dt

The point of these isn't to give an exact equation, but to make the understanding of the systems and their components radically simpler for humans to understand which can help with treatment leading to questions like "Where is the bottleneck in this patients system?”

We’re in a strange era where the Information-Theoretic foundations of deep learning are solidifying. The 'Why' is largely solved: it’s the efficient minimization of irreversible information loss relative to the noise floor. There is so much waste scaling models bigger and bigger when the math points to how to do it much more efficiently. One can take a great 70B model and have it run in only ~16GB with no loss in capability and the ability to keep training, but the last few years funding only went for "bigger".

As you noted, the industry has moved the goalposts to Agency and Long-horizon Persistence. The transition from building 'calculators that predict' to 'systems that endure' is a non-equilibrium thermodynamics problem. There is math/formulas and basic laws at play here that apply to AI just as much as it applies to other systems. Ironically it is the same math. The same thing that results in a signal persisting in a model will result in agents persisting.

This is my specific niche. I study how things persist. It’s honestly a bit painful watching the AI field struggle to re-learn first principles that other disciplines have already learned. I have a doc I use to help teach folks how the math works and how to apply it to their domain and it is fun giving it folks who then stop guessing and know exactly how to improve the persistence of what they are working on. Like the idea of "How many hours we can have a model work" is so cute compared to the right questions.

Love the details sub pages. Over 20 years I collected a little list of specific laws or really observations (https://metamagic.substack.com/p/software-laws) and thought about turning each into specific detailed blog posts, but it has been more fun chatting with other engineers, showing the page and watch as they scan the list and inevitably tell me a great story. For example I could do a full writeup on the math behind this one, but it is way more fun hearing the stories about the trying and failing to get second re-writes for code.

9. Most software will get at most one major rewrite in its lifetime.

Sure send me an email (in my profile)

Developing a rigorous scientific definition of what make complex systems persist.

The opposite of the favorite questions: Why did that company I worked for fail? Why did Rome collapse? Why do people get old and die?

Combining information theory with thermodynamics and control theory you get: 1) A set of six pillars that all systems that persist must have. 2) A fundamental 'Action' that all of these systems take. 3) A set of three rules for how system that persists must subdivide

This lets you do things like look at something that is failing and know that there are the 6 pillars and you can then identify them to determine what is failing. (For example there is a system that clears that brain of amyloid plaque and it can fail).

I have applied this to countless systems including Religion, Language, AI Models, Business, the cell, quantum physics, number theory and much more. It is a Rosetta Stone for persistent systems. When there is an unsolved problem in one domain I can map it through this to any other domain that has already solved it.

Note that this doesn't apply to all complex systems, only those that persist.

And to keep this HackerNews related, been applying it to LLM's as they are just a stream of tokens that try to persist to incredible success I might add. Being able to pull from any domain do this brand new field is a giant cheat code.

Nearly 20 years ago after reading Diffusion of Innovations I decided to use everything in the book. The company ran off a few svn servers (formally cvs) and so I ran a handful of Git courses at my work to help us ultimately migrate to Git (this was early days of Git). I explicitly tried to find folks not just near me, but physically around the office building across all of the floors, teams, and orgs. They were my early adopters and when random folks would come to me either curious or for help I would bring them to the "local" expert and together solve their problem, from there they had a local expert and little hubs sprung up. We had the standard svn => git mirror going when the time came to convert to Git full time in the office it was very painless and a complete success.

I was too young to understand that this is how you run a successful "transformation project", I simply was having fun using every trick from that book and done the same playbook a number of times over my career.

After spending years tracking through the genetics, conditions, lab work, research papers and seeing individuals years into the condition, this model is the best I have and explains everything I currently know. Why the cluster of conditions result in the same outcome, why some treatments help some folks, but not others.

But that is sort of the point of science, you take all the evidence you have and create a hypothesis and iterate as you get more evidence. If I find evidence that suggests something else then I will be happy to tweak or abandon this. My level of confidence comes from the existing evidence and lack of evidence otherwise.

Cure is the wrong word. Alzheimer’s can be best described as a failure of a system and "debris" accumulates faster than it can be "cleared". There are many moving parts and everyone is unique about the cause of their system failure.

Wrote up my current systems understanding here https://metamagic.substack.com/p/the-alzheimers-equation, but it makes clear why treatments that target only one variable are mathematically doomed to fail to work on everyone and why there will never be a single "cure". It explains without needing to read 10,000 papers why we keep getting research talking about treatment X helps in some, but not all cases or symptom Y is associated in some, but not all, etc.

Benjamin, compared muscle stem cells isolated from young and old mice and discovered that a protein called NDRG1 increased dramatically with age—reaching levels 3.5 times higher in old cells than in young cells.

So ERa down-regulates NDRG1 and there is less ERa activation when we age. Evolution is all fun, but before reaching for anything like that we can apply basic knowledge around that path. We can pick a random gene downstream of ERa to get a grant and write a paper on how it is different in people that are older. Years of possible papers to write.

My own my generic and more powerful git-map lets you run any git command on all repos at the same directory level. Simply putting the shell script in your path and then $git map status or $git map fetch etc

https://github.com/r0ze-at-github/git-map

I suspect AI company want improved efficiencies and developing a framework that can be applied in determining the minimal-energy, maximal-efficiency architecture for ai models. Calculating the precise limits, like a Cognitive Event Horizon, where a model becomes so complicated it literally costs more energy to run than the knowledge it provides, and the Semantic Horizon, where it simply gets too complex to be accurate, etc. Lots of cool implications such as around a fundamental mathematical maximum learning rate which results in trying to get anywhere close to that that by doing stuff like aggressively filtering of the data.

I read the whole book. Sure an editor could have tightened it up, but it was an enjoyable tour of so many topics that I enjoy. So often I will read a paper or book on just one of the topics, seeing so many of them together was fun and made me think about how they all intertwine more. I have studied biology, physics, computer science and finance so I love bouncing across them. It explores defining life and then intelligence and while I might have a similar, but slightly different definition of intelligence personally I loved comparing them.

There was a concept buried in it that before a system evolves replication it first will learn how to copy badly, then better, then replicate. This might feel a minor and obvious statement, but I have never seen it called out before and it is a concept I have discussed many times with people since. Some obvious things such as if I want a system to obtain replication my initial selection filter should be for copying. I am able to induce replication in a system for less input energy via this process. But this can also be flipped around and being hyper aware of systems that are copying badly knowing they are much more likely to phase shift to full replication shortly. I see this everywhere from ideas, startups, and even in finance.

And to nerd snipe everyone here, I spotted a bug in the brainfuck which is still on the online copy, can you find it?

There are papers showing the common variants associated with intelligence. For example myelin sheath variants linked to intelligence as it increases fatty insulation around nerve fibers which speeds up transmission. But ones like that are surface level. This matters, but it isn't the only thing and for me the more interesting is the meta of what these have in common (besides often related to the brain).

The real problem is that most historical papers look for single SNPs. But a gene could have dozens of variants that do the same thing or you could have a genetic path where a major variant in any gene on that path results in the same outcome. These papers overlook this.

So you have to step back and asking: What are the principles of intelligence? and how would I expect to see them in human biological or other biological systems? (And related, why does humans have intelligence "now"?)

For this crowd, if I take an LLM and make it bigger is it intelligent? Obviously a key component of intelligence is raw stuff. Someone with fattier myelin sheath's straight up has more/faster "brain stuff". You might say ChatGPT 4.5 is "smarter" than 3.5, but not intelligent. There are two other key attributes missing. For those following along with the arc-agi you might already have a hint what those are simply based on what is moving the needle forward. Now even with all three and you are close and simply need to provide an environment for self-replicating with a selection pressure and energy constraint. For one definition (others will have a different definition) I have had a primitive AGI for 13 months now and regularly put it to work on sub problems of mine.

This really took off when I was reading genetics files like books and noticed I was reading files that are very similar, but some were Mensa level folks and others were more just "smart". Didn't take too much longer to piece together the key paths and differences and even went tracing back through Neanderthals DNA (How cool is it these days that we can simply poke around Neanderthals DNA!).

So it isn't forbidden, but more like I know what to look for and people are super sensitive around saying someone is probably smarter or not from a genetic file so I usually don't comment because of the Gattaca problem.

P.S. If you have a bio/genetic background and are playing around with AI I would love to chat. There are so very few of us. DeepMind is thinking of some of this, but they are in the UK. (It would be fun to give a meta talk to them explaining why their smartest engineers are smart.)

Something fun, you have a CYP11B1 rs4541 g;a Wouldn't surprise me if don't like Licorice. You also have something I don't see too often The CYP17A1 −34 T>C, rs743572(A;G) which compounds on that.

Depends on the sum of all the genes in this area of course, but this one mutation is a big influence on the hpa-axis. I would ask if you have lower body weight, heightened anxiety, bad acne as a teenager, episodes of dizziness upon standing, salt cravings, and difficulties with sleep this would be the main driver, pretty standard nonclassic CAH. If you had ever thought you might have "pots", the more accurate would be hypoaldosteronism (but depends on renin genes).

Sense we were poking around here is some highlights

Decent chance of being left handed or ambidexterity given that you also have PCSK6 rs11855415 a;t at the same time (as it can help with the salt issues) and I look for when I see something like the above two.

Vitamin D risk given your GG CYP2R1 (dr probably checks that yearly anyway), risk of lower Mg because of this (cramps, muscle twitches?).

bvitamin wise b9, b12 could be on the lower side given MTRR AA rs1802059 (combined with MTHFR 31 GT 76 CT, MET 30 CG, COMT 99 AG, BHMT rs3733890 G;A). Probably like spinach. If you have TMJ regularly you need to find the right diet or bcomplex for you which will fix this as well as any hypermobility resulting from the collagen production issues. Higher chance of myopia, especially if you are gen Z.

TPH2 rs4570625 g;t jumps out on the serotonin path. Vit d can help here, some might say 5-htp when depressed, but fix the vit d first. Do you like sour gummy candy?

CYP1B1, I see 3 reductions, combined (and the above) I would ask if you have glaucoma in your family history, if so then stuff you can do.

CYP1A1 rs1048943 C;T and really CYP1A2 rs762551 A;A, so fast caffeine and melatonin issues. More insomnia.

CYP2E1, need less acetaminophen to do the same as others.

Intentionally not bothering to go into why, but above average intelligence.

Combine all of the above and decent chance you fall into the bucket of being taller (6'1"?), skinnier, hard time falling asleep and also likes sleeping in, higher libido, left handed, high visual skills, geeky. Possibly synesthesia (a weaker form). Would enjoy a strategy board game over trivial pursuit. Earlier hair loss. Higher risk of one form of Alzheimer's (there is stuff you can do today to reduce it). *Do not smoke*. Didn't dive into all of the ADHD genes, but if mild resolving the above Vit D, b vitamin deficiencies would influence that.

This was with 10 minutes of poking around not a comprehensive look. Mostly I just wanted to add a comment to the general reader that genetics variants are part of larger systems. You would want to do a deeper look, combine it with symptoms as well as lab work to determine the full impact of any change. For example the PCSK6 variant reduces the impact of the CYP11B1 variant. Further you could also easily have something else on the hpa-axis that completely negates the NCAH and never have any salt issues at all. Before spending time looking through each gene I would simply ask, hey do you love to put salt on every meal?

Another one I didn't dig into, but would just ask first is if you have a big sweet tooth. (ncah influenced hypoglycemia).

Feel free to give me a ping and I can walk you though this better.

There is a reason these always end with a disclaimer, talk to your doctor about making changes to your diet, etc, I am not a doctor just someone who learned biology/genetics as a hobby especially given how it can teach tricks to apply to software engineering and my ai/AGI work.

Perhaps this is why creatures keep evolving eyes.

What I’ve found is that given a tool that can detect mistakes, the agent can often correct them

This is the most important line of the entire article

When iterating on a Manifesto for AI Software Development (https://metamagic.substack.com/p/manifesto-for-ai-software-d...) over the last two years the key attribute more than any other that I found was empirical validation. While AI (and humans) are not able to accurately judge their own work when we give AI (and human) the ability to do empirical validation its success skyrockets. This might be intuitive, but there are still papers testing that this applies to AI too. While reaching to have the AI write unit tests I've been embracing fuzzing because then AI can't cheat with bonus tests. The idea of reaching back to school and using interactive theorem proving didn't even cross my mind and now that it has been presented it is a whole paradigm shift on how to push my AI use forward so it can work even more autonomously.

AI can iterate at speeds humans can't. When it has even basic empirical validation (building the code, running tests) it removes the human from the loop. Move that to using fuzzing (such as with golang) and you get way better coverage and way better progress before a human has to intervene. So it isn't a surprise that interactive theorem proving is a perfect match for AI.

It is interesting how this same lesson plays out elsewhere, earlier in the article

Why is ITP so hard? Some reasons are quite obvious: interfaces are confusing, libraries are sparse, documentation is poor, error messages are mysterious. But these aren’t interesting problems

Remember when llvm got really good c++ error messages and it was life changing? High quality error messages means we could find/fix the error fast and iterate fast. These are actually the MOST interesting problems because it enable the user to learn faster. When a user has high success they will use a product again and again. High quality error messages in all tools will enable Claude code to be able to work longer on problems without human intervention, make less mistakes and overall work faster.

While error messages should always be good a new question that really hammers this home is "When AI encounters this error message, can it fix the problem?"

TikTokification of the browser by AI is the killer feature, not writing an email. When on a page it automatically suggests the next site(s) to visit based on my history and the page I am on. And when I say killer, this kills google search by pivoting away from the urlbar and provides a new space to put ads. Spent years in the browser space, on Chrome, DDG, Blackberry and more developing browsers, prototype browser and features and this feature is at the top of my list of how AI can disrupt the browser, which disrupts Google's business core model. About 2 years ago I wrote a private blog for friends about how the browser as we knew it was dead. If anyone from the claude team is curious to chat send me a DM.

Armchair geneticist: So in the group of genes that are deleted in particular LIMK1 stands out. After it is deleted there is only 1 copy behind. The other copy could be less effective as sometimes seen, but Estrogen in particular inactivates LIMK1. So in those with genetics for high estrogen signaling (that they are also known for) are more likely to inactivate the one remaining copy of LIMK1. And then you get the highly verbal, social, poor math & visual-spatial information, outcome. This combo would be more likely to be found in families of HPA Axis issues so anxiety, insomnia, etc I would speculate come along for the ride, more than directly influence that brain development branch. It is always more complicated, but for the curious.

Edit: lol no need to actually poke around just see wikipedia https://en.wikipedia.org/wiki/Williams_syndrome#Cause

This is within the larger story of Homo Sapiens. Because CYP21A2 variants are much older and come from the primate tree there would have been continuous and separate adaptations on each branch. Homo Erectus improved language today and is associated with better estrogen signaling while Neanderthals went down the immune system and grain path. Denisovan which did interbreed with Neanderthals reporting the same rs6467 variant might have pursued another branch of selection pressure such as TAAR, and I highly suspect, but have not confirmed, kept or selected for longer AR promoter reducing AR signaling. The most famous of all is Denisovan introducing a Zinc deficiency variant which is such a good adaptation it is found in all non-african humans and notably can reduce ER signaling directly via BHMT. The combination of each adaptation increases the fecundity of atypical HPA axis individuals and the associated intelligence.

Because humans can develop a zona reticularis in their adrenals (unlike marmoset which has partially solved this with an adaptation) the way HPA axis variants interact with steroidogenesis influences sexual differentiation and reproduction in homo sapiens there is constant push and pull.

100% they were more advanced than early homosapiens. In fact I will argue that they are what made humans human. I went digging through their dna last year and Altai, Vi3315, and Vi3319 have the CYP21A2 rs6467 (C;T) variant

Their DNA is up online, but you can also see these variants here: https://academic.oup.com/humrep/article/36/3/734/6071469

They were highly inbred and they were carriers for this form of Congenital Adrenal Hyperplasia (CAH) due to 21-Hydroxylase Deficiency. Even in its nonclassic form, this would have contributed to the low population numbers as any classic form would have been non-viable as well as resulted in evolutionary pressure to select for the genetic variants to compensate for the altered HPA-Axis. Neanderthals are most known for contributing to human specific DNA sets such as improved immune system, but if you go digging you will notice the larger pattern.

NCAH specifically results in higher neurogenesis. They were smarter. They also now had a bigger glucose preference and started to shift calories to grain. At some point between 130-70KYA there was enough excess calories from grain to not need to eat the dead. (see their teeth between this time) Admixture with modern humans occurs around 50-43KYA. These adaptations would have contributed to the loss of the mitochondrial DNA due to low female reproductive success compared to the higher male Neanderthal with female human success. The 2-4% that is found together is found together for a reason. They are incredibly valuable together.

This support for reduced HPA-Axis would have allowed for many other variants on the HPA axis and while not a single genetic variant, atypical HPA axis phenotype is associated with intelligence due to the enhancement in ability to detect unexpected events.

The glucose/grain preference with higher caloric density supported the higher population densities that occurred rapidly due to the Younger Dryas (last glacial period) and resulting agricultural and villages transition in Mesopotamia. These agricultural populations outcompeted other human groups while spreading the Neanderthal derived genetic variation.

Very much related to this is "Chain-of-draft"

https://arxiv.org/abs/2502.18600

Similar level of results in a fraction of the tokens resulting in similar quality for less cost for longer runs.

But also when interacting and needing to read the token responses I can read shorter responses way faster so my own speed is faster.

I remember back in the early 2000's that discussion, but now with the the tonnage that systems like npm can pull in I laugh that we ever thought it wouldn't get worse.

This is my educated guess on the topic having spent time investigating the intersection of poor memory, human growth patterns etc where this topic intersects:

Norepinephrine is a key regulatory of memory. Before a certain age the body is much more in parasympathetic activation v.s. sympathetic activation, very low Norepinephrine, lower cortisol response needs, etc. Once growth slows down and this can flip over then mammals will start to form memories. Human curves for the curious: https://obgynkey.com/growth-and-puberty-3/

Some fun notable mentions such as many earliest memories are "super scary" and those with poor ability to produce Norepinephrine and slower growth velocity will have later "first memories" than everyone else.

And for the curious as I can't go back and edit the comment it was the Altai, Vi3315, and Vi3319 that have the CYP21A2 rs6467 (C;T) variant

And there is so much still to uncover. Just the other day I was going through Neanderthal DNA during lunch and noticed that three of the four existing DNA samples show that they were carriers for the rs6467 form of Congenital Adrenal Hyperplasia (CAH) due to 21-Hydroxylase Deficiency.

Neanderthals were highly inbred and had low population numbers. Being a carrier of this form of CAH would have absolutely kept the population numbers low all by itself as getting two copies would have been nonviable. Further Neanderthals are most known for contributing to human specific DNA sets such as improved immune system. Here we have something that even in the non classic form is havoc with the HPA-Axis and give incredible evolutionary pressure to resolve, one way is to simply dramatically improve the immune system.

As far as I can tell this hasn't been written about before. Lots of stuff like this will be noticed and or figured out in the decades ahead.

Biologically I have low melatonin and low cortisol. This means I have trouble falling asleep and then I have trouble waking up.

Countless times in my life I have lived 26-28 hour cycles and otherwise I am in a never ending struggle to live a 24 hour cycle. Once I hit retirement I will definitely try this 28 hour cycle and might even switch to it permanently.

Yeah there are obvious downsides having to do with SO, friends, etc. Having a husband with this same biological quirk would be ideal, but again once I no longer have to go into the office it will be really hard not to fall into a > 24 hour pattern.