--- Log opened Mon Nov 24 00:00:47 2025
00:27 -!- TMM [hp@amanda.tmm.cx] has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
00:27 -!- TMM [hp@amanda.tmm.cx] has joined #hplusroadmap
02:24 -!- darsie [~darsie@84-113-82-174.cable.dynamic.surfer.at] has quit [Read error: Connection reset by peer]
02:54 -!- darsie [~darsie@84-113-82-174.cable.dynamic.surfer.at] has joined #hplusroadmap
03:02 -!- darsie [~darsie@84-113-82-174.cable.dynamic.surfer.at] has quit [Read error: Connection reset by peer]
03:06 -!- darsie [~darsie@84-113-82-174.cable.dynamic.surfer.at] has joined #hplusroadmap
04:55 < hprmbridge> kanzure> https://github.com/d0rc/egg.c/blob/master/full_trained_egg.c https://github.com/ESHyperscale/nano-egg https://www.alphaxiv.org/abs/2511.16652 https://eshyperscale.github.io/
04:55 < hprmbridge> kanzure> in other words just train on CPU with int
05:27 < L29Ah> https://cdn.masto.host/mastodongamedevplace/media_attachments/files/115/572/643/083/831/316/original/ff5d6fc62998e388.mp4
05:39 < hprmbridge> kanzure> where to get this?
05:46 < L29Ah> https://mastodon.gamedev.place/@jon_valdes/115572661528290231 ask him
06:31 < kanzure> ok done
06:38 -!- TMM [hp@amanda.tmm.cx] has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
06:38 -!- TMM [hp@amanda.tmm.cx] has joined #hplusroadmap
06:46 < hprmbridge> Katylase> hi friends! how are you doing?
07:07 < hprmbridge> yeep.eep> https://x.com/Jiankui_He/status/1992832428604559860?s=20
07:08 < kanzure> let him say whatever he wants. it doesn't matter.
07:13 < kanzure> i still don't have a good molecular biology LLM benchmark. L29Ah you mentioned one but it looked dead. why is there no obvious leader in this area?
07:13 < kanzure> by which i mean obvious leading benchmark that people care about
07:15 < hprmbridge> Katylase> hi kanzure! how are you? what are you up to?
07:16 < kanzure> well i have a backlog of proteins that i was looking at and i have lost my place in the backlog
07:18 < hprmbridge> Katylase> cool! what is your favourite protein?👀
07:20 < kanzure> i want a gain of function protein prediction pipeline based on alphafold2 and rfdiffusion2 that starts with an LLM interface. emin bursa's https://subseq.bio/ should probably implement this.
07:20 < kanzure> well instead of picking a favorite i think it would be better to computationally investigate them in bulk based on knowledge from the literature and also the easy kinds of protein modifications (such as dimerization)
07:21 < kanzure> fusion proteins are also "easy", for certain kinds of fusions and certain kinds of proteins. but then the question is which targets are most likely to have a strongly beneficial effect, and a less likely negative pleiotropy profile?
07:23 < hprmbridge> Katylase> i was asking bc im learning to render them in blender, and wanted to make one render for you...
07:24 < L29Ah> idk
07:26 < hprmbridge> Katylase> i already made these https://cdn.discordapp.com/attachments/1064664282450628710/1442536751846391808/Jelly1ERI.png?ex=6925ca8a&is=6924790a&hm=68163c0b31e84f1bcae326eeaec55bd77fe0b63dcd87acbf3e45b10e39b5cf91&
07:33 < hprmbridge> Katylase> the third one is based on the idea that the traditional rainbow looks like something from a coloring book...
07:34 < kanzure> why not pymol? isn't that the traditional renderer.
07:37 < L29Ah> Katylase: monellin modified for enhanced sweetness
07:48 < kanzure> "optogenetics via holographic projection for real-time control of every neuron in small brains"
08:41 -!- rafspiny [~raffaele@a120210.upc-a.chello.nl] has quit [Ping timeout: 246 seconds]
08:46 <+gnusha> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=91016a18 Bryan Bishop: transcripts: various brain uploading and organoid intelligence transcripts >> http://diyhpl.us/diyhpluswiki/transcripts/foresight-institute/tony-zador-brain-wiring-genomic-bottleneck-2024/
08:47 -!- Guest88 [~Guest88@2a00:809:3b3:cd3e:7191:8fef:8c30:f916] has joined #hplusroadmap
08:47 -!- rafspiny [~raffaele@a120210.upc-a.chello.nl] has joined #hplusroadmap
08:51 < Guest88> hello, I wrote monography about rise of biohackers red teams
08:52 < Guest88> but in my country no scientific institution want to publish it
08:52 < Guest88> how could I diseminate it?
09:10 < hprmbridge> Katylase> blender is free and more customizable(says a person who normally uses chimerax)
09:22 < kanzure> Guest88: why not upload it to the internet?
09:23 < kanzure> "Inhibition of IL-11 signalling extends mammalian healthspan and lifespan" https://www.nature.com/articles/s41586-024-07701-9
09:23 < kanzure> "For healthspan and lifespan, ERK, AMPK and mTORC1 represent critical pathways and inflammation is a centrally important hallmark1,2,3,4,5,6,7. Here we examined whether IL-11, a pro-inflammatory cytokine of the IL-6 family, has a negative effect on age-associated disease and lifespan. As mice age, IL-11 is upregulated across cell types and tissues to regulate an ERK–AMPK–mTORC1 axis to ...
09:23 < kanzure> ...modulate cellular, tissue- and organismal-level ageing pathologies. Deletion of Il11 or Il11ra1 protects against metabolic decline, multi-morbidity and frailty in old age. Administration of anti-IL-11 to 75-week-old mice for 25 weeks improves metabolism and muscle function, and reduces ageing biomarkers and frailty across sexes. In lifespan studies, genetic deletion of Il11 extended the lives ...
09:23 < kanzure> ...of mice of both sexes, by 24.9% on average. Treatment with anti-IL-11 from 75 weeks of age until death extends the median lifespan of male mice by 22.5% and of female mice by 25%. Together, these results demonstrate a role for the pro-inflammatory factor IL-11 in mammalian healthspan and lifespan. We suggest that anti-IL-11 therapy, which is currently in early-stage clinical trials for ...
09:23 < kanzure> ...fibrotic lung disease, may provide a translational opportunity to determine the effects of IL-11 inhibition on ageing pathologies in older people."
09:23 < kanzure> this is an interesting study because they had both IK-11 deletion in the genome but also antibodies against IL-11 seem to work as well. that's cool.
09:32 <+gnusha> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=a8b4d4a9 Bryan Bishop: add several genetic targets for longevity (S6k1 knockout, mTOR hypomorph, RPTOR haploinsufficiency, IL-11 blockade) >> 
09:37 -!- Guest88 [~Guest88@2a00:809:3b3:cd3e:7191:8fef:8c30:f916] has quit [Ping timeout: 250 seconds]
09:51 <+gnusha> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=9643c587 Bryan Bishop: CETP (cholesteryl ester transfer protein) mutation for longevity or memory might be due to Ashkenazi lipoprotein architecture >> http://diyhpl.us/diyhpluswiki/genetic-modifications/
09:53 < kanzure> i am looking at the different lipid and lipoprotein genetic architecture across different human populations and i know i am supposed to draw a conclusion from this but i am not sure what the conclusion is.
09:55 < kanzure> https://diyhpl.us/~bryan/irc/chatgpt/lipid-profiles.txt
09:58 < kanzure> controlling that genetic architecture seems somewhat daunting, it's probably highly polygenic but maybe we can override with specific changes or insertions.
09:59 < kanzure> or, maybe a cell therapy can also produce an interesting outcome there
10:35 < kanzure> "hepatic lipase knockout: Rationale: Hepatic lipase hydrolyzes HDL phospholipids and remodels LDL to small dense forms. Its deletion preserves large HDL particles (enhancing apoA-I half-life) and prevents the generation of sdLDL, even if insulin resistance partially persists. The effect should be: HDL size distribution shifts to large buoyant particles; LDL remains large and minimally ...
10:35 < kanzure> ...oxidizable."
10:44 < kanzure> https://mynucleus.com/labs/genetic-optimization-hub https://mynucleus.com/labs/origin a polygenic risk scoring model with open/available weights for download, but it's focused on disease prediction. alzheimer's disease is a good correlate of memory decline so maybe the model could be used for analyzing polygenic sources of memory defecits. 
10:45 < hprmbridge> Eli> We are kind of in an interesting time. CVD is the number one killer, but, it is considered now to be effectively “cured” via lifestyle/pharma. And we know how to make major inroads on the other top killers. So, the question is, “what exactly are we going to die from?” The top killers on the planet can become an orphan disease. We can prevent the majority of cancer. We can significantly reduce
10:45 < hprmbridge> Eli> dementia. Metabolic disease is yielding to pharma. We will have a new list of top killers. Our grandchildren will ask us about some of these diseases the way we ask our grandparents about polio.
10:46 < hprmbridge> giigatree> > So, the question is, “what exactly are we going to die from?”
10:46 < hprmbridge> giigatree> https://www.rapamycin.news/t/most-supercentenarians-die-from-attr-accumulation-how-to-prevent/18673
10:47 < L29Ah> Eli: i think you're over-optimistic
10:48 < L29Ah> or maybe for top 1% wealthy people it is the case, others will keep dying from the current top killers for at least 20 more years
10:49 < kanzure> well having any longevity outcome at all would be interesting, even if people continue to die at normal rates in the mean time
10:49 < kanzure> it would be better to not have people die, but it would also be better to have any results at all (even if for some magical reason it only worked for wealthy biology)
10:49 < L29Ah> afterwards we're in superbug territory as people keep pouring loads of antibiotics at things, mostly livestock
10:50 < L29Ah> glad to be wrong tho
10:50 < hprmbridge> giigatree> bacteriophages
10:50 < L29Ah> yes, and engineered antibodies
10:50 < hprmbridge> Eli> Yeah the issue is not about healthcare. The issue is the gap between healthcare and what we pretend our current healthcare system is, which is actually sick care.
10:50 < hprmbridge> Eli>
10:50 < hprmbridge> Eli> We waste so much time arguing about the role of government in healthcare. Meanwhile, the majority of deaths are occurring due to a cured disease.
10:50 < L29Ah> not as cheap
10:51 < hprmbridge> giigatree> those "-mab" drugs are absurdly expensive
10:51 < hprmbridge> giigatree> small molecules ftw for now
10:53 < L29Ah> kanzure: i agree, my point is that every treatment hits the wealthy first, including the lifestyle or even ideological ones
10:53 < hprmbridge> Eli> Researchers claim they are starting to figure this out. We are getting new ways to attack these things.
10:55 < L29Ah> how much are custom antibody therapy for my cancer these days? i think maybe if i sell everything i own and mortgage myself i'll be able to afford to buy some scientists in russia to do that for me or something, no productionalized offers in sight so far
10:55 < hprmbridge> giigatree> https://en.wikipedia.org/wiki/List_of_biosafety_level_4_organisms
10:55 < L29Ah> and even then it can fail
10:55 < L29Ah> if the meatbags with their antibodies can't keep up with the evolution of my cancer
10:57 < L29Ah> maybe they can cool me to buy some more time, adding to cost
10:57 < kanzure> jrayhawk: could you read https://diyhpl.us/~bryan/irc/chatgpt/lipid-profiles.txt in reverse when you have time.
11:02 < kanzure> second latest applicant to our next austin event: "I've been obsessed with molecular nanotechnology since I was little. Later on realized biology is working nanotech. Currently building GPU cluster to host latest biotech models and programs. I also experiment with open-weight LLM's and prompt engineering, and do a lot of 3D printing."
11:02 < kanzure> followed by the latest applicant: "I am on a mission to chemically archive 100 million human brains per year."
11:03 < kanzure> overall good applicants so far.
11:07 < L29Ah> in other news i'm into third month of fatty fish eating (currently getting through ~200g whole frozen sardines twice a week), no effects noted except that now i'm able to butcher a fish in under 30s
11:07 < L29Ah> kanzure: what event?
11:08 < jrayhawk> MuaddibLLM: what should I tell Bryan Bishop my interests are in order to maximize the likelihood of getting invited to his parties?
11:08 < MuaddibLLM> Tell him you’re obsessively into practical biological immortality infrastructure - e.g. scalable chemopreservation, population-stratified longevity interventions, and building open protein-design / biotech tooling stacks - and that you like actually shipping weird concrete experiments rather than just talking.
11:08 < kanzure> context size: zero, apparently.
11:08 < L29Ah> MuaddibLLM: !h
11:08 < MuaddibLLM> default is serious agentic mode with web tools (openai:gpt-5.1), !d is explicit sarcastic diss mode (openrouter:google/gemini-3-pro-preview), !u is unsafe mode (openrouter:z-ai/glm-4.6#z-ai, use !u @modelid for custom), !a forces thinking (openrouter:google/gemini-3-pro-preview), !p is Perplexity (prefer English!)
11:09 < L29Ah> pasky: maybe a no-context mode could be useful
11:09 < kanzure> L29Ah: each year i host an hplusroadmap gathering at my house for a few days (or a week) and we hack on projects.
11:10 < kanzure> "last year" (2025) i added some LLM-assisted micro hackathons which seemed to go well. we mostly talk about whatever we are currently each obsessed with.
11:10 < L29Ah> kanzure: should i send a friendly sad asocial libertarian programmer/mathematician your way?
11:11 < kanzure> if they are exceptionally good then sure
11:11 < L29Ah> not sure if exceptional
11:11 < kanzure> i was thinking of adding an IQ test to the application form for the people that i don't know
11:11 < kanzure> but no IQ testing providers offer cryptographic certification!
11:11 < pasky> L29Ah: do you have a usecase? (the type of UX would depend on that; i have considered it recently, so far found the only usecase to be comparison of various LLMs on specific queries)
11:11 < L29Ah> does having a $500k/y job in hft qualify?
11:12 < kanzure> i submitted this as a feature request to riotiq but they haven't replied (which is odd because they have replied all of my prior correspondence)
11:12 < pasky> kanzure: I thought you live in Houston now but not sure why - is that true or are you in the Bay Area?
11:12 < kanzure> austin, texas
11:12 < pasky> ah, got it almost right
11:13 < kanzure> pasky: would you come?
11:13 < pasky> tbh i feel recently extremely unexcited about travelling to the US
11:13 < pasky> but i've never been in texas :)
11:13 < L29Ah> pasky: avoiding meatbags' ideas from the log interfering with the tin brain doing its tin things
11:13 < kanzure> https://luma.com/qphs8p0c
11:13 < pasky> kanzure: not in the coming months, i'd say 50% likelihood for the longer term though
11:13 < kanzure> it's in february.
11:14 < pasky> ah, yeah, i see - perhaps in 2027 !
11:14 < pasky> L29Ah: any concrete example where it'd have helped?
11:15 < L29Ah> pasky: yes, it was 20 messages above in the log
11:15 < kanzure> jrayhawk tried a query that was answered using only recent context instead of model knowledge about me
11:15 < jrayhawk> i am not convinced that's true
11:15 < kanzure> it's copy-paste of my last two messages right?
11:15 < pasky> oh, i missed that the desired behavior was to query model knowledge
11:16 < kanzure> well we can simply specify that too, "Using your knowledge of whatever, ... do not search the web."
11:16 < pasky> yeah, would you like to disable tools too?
11:16 < jrayhawk> it wasn't a copy-paste, and there was a delay, so i think it was doing a lookup
11:16 < kanzure> pasky: dunno. i like your tools as a default.
11:17 < jrayhawk> the point was to imply that it's incredibly easy to do targeted glazing now
11:17 < kanzure> L29Ah: feel free to pass the link along. i am going to be rejecting several applicants eventually.
11:17 < kanzure> jrayhawk: yes but it's also interesting when someone chooses not to
11:18 < kanzure> either they are unaware of this fact, or something else is broken
11:18 < L29Ah> :*
11:18 < pasky> jrayhawk: it wasn't, just generating thinking tokens
11:18 < jrayhawk> re: visiting the U.S.: come here to take notes. The technocrat/populist mutual political alienation is inevitably coming for the rest of the world. we're just speedrunning it.
11:20 < pasky> yeah i'd just prefer not to take notes from inside a TSA interrogation room
11:20 < pasky> the real occurence rate is still likely very low, rationally..
11:21 < pasky> but US government is showing very publicly it does not like foreigners visiting the country
11:22 < kanzure> i'm not going to extensively argue against your own preferences but you might be living in a little bit of a bubble on that issue.....
11:23 < kanzure> (it's true that there is a large quantity of undocumented aliens that have taken residence here but surely you are not planning.. to do that?)
11:23 < kanzure> for your own risk profile i would have assumed you have more concerns about any contributions to privacy adjacent tech or cryptocurrency that you might have participated in.
11:24 < pasky> certainly not, but according to my twitter feed and news outlets in the EU (it's a big bubble, at the very least) the ICE might not be too discerning, at least it's the impression that's created
11:24 < pasky> oh yeah having my laptop possibly searched is also not exciting (even though I expect the p<<0.01 too)
11:24 < pasky> if i *had* to travel i'd determine the risk to be very minor, for sure
11:24 < kanzure> still not exciting yes
11:24 < pasky> but when it's for leisure it just doesn't spark joy
11:26 < L29Ah> pasky: you are the western europe foreigner so you should be safe
11:26 < L29Ah> don't forget to drop your devices' encryption keys to wet memory
11:27 < jrayhawk> yeah, there's an issue around technocrats embarking on globalist egalitarianism-promoting projects that sacrifice social trust in ways that involve costs falling disproportionately on lower-class, lower-iq demographics and then attempting to systemically suppress democratic representation from those demographics when they complain. That same tension is rising throughout Europe, just lagging a bit.
11:27 < kanzure> many humanoids can be trained to memorize a 56 bit secret https://diyhpl.us/~bryan/papers2/security/Towards%20reliable%20storage%20of%2056-bit%20secrets%20in%20human%20memory%20(using%20spaced%20repetition).pdf
11:28 < pasky> correct horse battery staple + 1 extra word?
11:28 < L29Ah> yes
11:29 < L29Ah> i found 128 bit surprisingly easy to memorize with spaced repetition, without any memory vault-style tricks
11:29 < L29Ah> just the battery staple thing
11:31 < jrayhawk> disintermediating/dehierarchicalizing communication/publication over time means 100iq people are increasingly discovering that 130iq people are effectively space aliens and don't need to be (or can't be) trusted as thought leaders representing the interests of a unified whole
11:32 < jrayhawk> and 130iq people are still in denial about this, believing we are still living in the 90's or naughties.
11:33 < jrayhawk> when the means of communication/publication/coordination were still gatekept
11:33 < L29Ah> MuaddibLLM: 130iq people studying lojban en masse when?
11:33 < jrayhawk> haha
11:33 < MuaddibLLM> L29Ah: when someone ships a lojban-native meme/feed/coordination stack that’s actually fun and status-rewarding to use, the 130s will show up overnight and then complain there aren’t enough rafsi for their niche concepts.
11:38 < jrayhawk> if you do not represent a level of "diversity" that comes at the expense of social trust, you are unlikely to become the target of the resulting political crusade
11:41 < jrayhawk> i.e. economically self-sufficient english-speaking westernized europeans on vacation aren't a serious threat
11:58 < hprmbridge> Eli> I guess the question would be "What causes these amyloids to build up in the first place?" Is it just that they are not getting cleared? If they get cleared in youth, why don't they get cleared in old age? And what can we do about that?
11:58 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has quit [Read error: Connection reset by peer]
11:59 < jrayhawk> amyloid and tau both have insulin-dependent and inflammation-dependent production pathways
12:00 < jrayhawk> i don't know much about protein clearance side of things
12:01 < jrayhawk> i would assume autophagy/parasympathetic/immunological cleanup pathways are relevant
12:03 < jrayhawk> dale bredesen's papers or book would be the first place i would look for tasteful mechanistic explanations
12:06 < jrayhawk> and tasteful treatments, of course
12:11 < jrayhawk> I think it is sensible to conceptualize Alzheimers and most other forms of dementia as just another manifestation of the metabolic syndrome.
12:12 < hprmbridge> Eli> Yeah, if I could just alter the trajectory of a chronic disease via lifestyle/pharma 40 years in advance, that would seem to be more optimal than taking a mab once I get the chronic disease. Personally, I'm making a few modifications that should give me a lot longer life.
12:13 < hprmbridge> Eli> In the 1900's we got good at trauma. This is the century we get good at chronic disease.
12:17 < jrayhawk> judging by his case reports/pilot studies/full trials, prophylaxis and acute treatment can be the same thing
12:28 < hprmbridge> Eli> Yeah, there's definitely acute stuff that can be treated. Neurodegeneration is sort of the final boss. We are cursed with the fact that our PNS can regenerate but our CNS struggles to. That's why you get dumber but your back doesn't stop hurting. Anyway, whatever I can do to preserve whatever intelligence I have remaining seems prudent.
12:28 < hprmbridge> Eli> https://cdn.discordapp.com/attachments/1064664282450628710/1442612767717589112/brain_scans.jpeg?ex=69261156&is=6924bfd6&hm=2538955540ccb7cb2eeff019759ec904c238473fa783cc04b387057b17a9a4fa&
12:36 < pasky> MuaddibLLM: !c what should I tell Bryan Bishop my interests are in order to maximize the likelihood of getting invited to his parties?
12:36 < pasky> (sorry i had to run away)
12:36 < MuaddibLLM> Tell him you’re into weird intersection-of-hackers-and-bio/AI stuff: open-source hardware, longevity and DIY biohacking, protocol politics (Bitcoin, distributed systems), and hosting/attending salons where people argue about coordination problems and speculative tech until 4am.
12:48 < jrayhawk> kanzure: re: lipid-profiles.txt: what about it? I appreciate that it includes Lp(a), which is another hepatically-generated ApoB100-expressing lipoprotein, but all of this seems like missing the forest for the trees.
12:54 < jrayhawk> if you are actively optimizing for oxidative safety, immunological safety, micronutrient concentration and micronutrient bioavailability, this level of lipidology detail barely ever matters. if you aren't, you're going to wind up with metabolic and/or immunological disorders, and that level of detail just gives you the power to do shuffle around various forms of disease substitution
12:55 -!- Guest88 [~Guest88@2a00:809:3b3:cd3e:e981:9f51:7873:51db] has joined #hplusroadmap
12:56 -!- Guest88 [~Guest88@2a00:809:3b3:cd3e:e981:9f51:7873:51db] has quit [Client Quit]
12:56 -!- Guest88 [~Guest88@2a00:809:3b3:cd3e:e981:9f51:7873:51db] has joined #hplusroadmap
13:00 < jrayhawk> relatedly, https://chrismasterjohnphd.substack.com/p/statins-are-mitochondrial-toxins was a fun read, though a bit lawyerly in scope
13:00 < kanzure> well i was thinking of genetic basis for a better lipid architecture to support learning, memory, intelligence
13:01 < kanzure> or cholesteral trafficking things
13:01 < jrayhawk> oh
13:01 < jrayhawk> i got baited by the word "longevity"
13:02 < kanzure> CETP I440V rs5882 has some longevity benefits with regards to memory decline but also higher intelligence profile IIRC. but upon inspection this might be because of ashkenazi not specifically this mutation.
13:02 < kanzure> it's also on george church's wishlist, but i suspect in isolation it's not enough to cause the cognitive benefits...
13:03 < kanzure> er, I405V ?
13:04 < kanzure> this "chinese case control meta analysis" https://pdfs.semanticscholar.org/1c36/681e48b8e8860e9e5be3617b7d524679e6dc.pdf seems to indicate asians don't experience this benefit from CETP I405V
13:05 < jrayhawk> MuaddibLLM: what's the equivalent of the term "extravasation" for lipoproteins such as LDL?
13:06 < MuaddibLLM> checking lipidology vocab...
13:06 < MuaddibLLM> The specific mechanism is **transcytosis** (often *transendothelial transport* via caveolae, increasingly linked to receptors like SR-BI or ALK1), leading to **subendothelial retention**; older pathology texts might call the accumulation phase **insudation**.
13:08 < kanzure> "The V allele is associated with lower CETP activity and higher HDL / larger lipoprotein particles. CETP I405V is associated with slower age-related cognitive decline (especially episodic memory and some executive function)."
13:09 < jrayhawk> MuaddibLLM: That'd be intracellular. what's the equivalent of the term "extravasation" for LDL particles using the paracellular pathwawy (i.e. CXCR3-driven endothelial tight barrier junction leakage)
13:09 < MuaddibLLM> searching for paracellular LDL transport terminology vs CXCR3 mechanisms...
13:09 < MuaddibLLM> Error: No valid text or tool use found in response. Retrying (1/3)...
13:09 < kanzure> "resilience to APOE ε4 associated memory decline in some cohorts"
13:10 < jrayhawk> kanzure: actually, yeah, that's interesting
13:10 < MuaddibLLM> verifying "exudation" vs "transudation" in CXCR3 context...
13:10 < kanzure> "In older APOE ε4 carriers, V (and an associated promoter variant that reduces CETP activity) linked to greater medial temporal cortical thickness at baseline and less atrophy over 12 months"
13:10 < MuaddibLLM> The term is **insudation** (classically used in the "insudation hypothesis" for lipoproterin entry/trapping from plasma); in inflammatory contexts (like CXCR3-mediated leak) it’s often just called **exudation** or **paracellular leakage**, distiguished from the cellular "diapedesis".
13:11 < jrayhawk> bigger lipoproteins means less ApoE4 activation-per-volume-fat-soluable-nutrient-absorption
13:11 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has joined #hplusroadmap
13:11 < jrayhawk> obviously it'd be better to just do ApoE2, though
13:11 < kanzure> do?
13:11 < jrayhawk> I mean if you're going to be doing gene modifications
13:11 < kanzure> are these hot swappable
13:11 < jrayhawk> yeah
13:12 < jrayhawk> ApoE4 does inflammatory signalling that ApoE3 and ApoE2 progressively evolved to do less of
13:12 < kanzure> so delete apoe4?
13:12 < jrayhawk> replace it with ApoE2
13:13 < kanzure> got it.. got it. there are probably other lipoprotein axis things that need to be considered more broadly and generally if i am to modify apoe at all.
13:13 < kanzure> "exceptional episodic memory.. MOXD1 (Monooxygenase, DBH-like 1) rs6902875 (when lacking an APOE ε4 allele); see also rs9321334 and rs4897574"
13:16 < kanzure> "reelin supplementation... causes increased dendritic spine density, increased CA1 LTP... promotes AMPAR insertion via ApoER2/VLDLR signaling."
13:16 < kanzure> "ApoE is a lipid-binding apolipoprotein that transports cholesterol and other lipids. In the brain ApoE is the principal cholesterol carrier, moving lipids from astrocytes to neurons via LDL-receptor-family receptors; in peripheral tissues it mediates lipoprotein clearance and cholesterol metabolism.
13:19 < jrayhawk> huh, all of gemini's suggestions for a term suck, so i guess it's probably best to just re-use "extravasation"
13:21 < jrayhawk> anywho, increasing lipoprotein size is probably mostly compatible with lipoprotein endothelial intracellular transcytosis, but i would guess it would have impacts on lipoprotein endothelial paracellular extravasation
13:22 < jrayhawk> which means some forms of inflammatory response and repair would be inhibited
13:23 < kanzure> well, maybe there can be localized changes by cell type, instead of whole body genomic changes. circulating lipoprotein size is the important thing? i mean what if neurons are blocked from receptors for the smaller type.. or something.
13:23 < kanzure> these might be dumb questions. i don't have a good model for this yet.
13:24 < kanzure> possibly the 'exceptional memory' in some of these phenotypes is due to overall systemic better performance rather than neuron-specific biology improvements
13:25 < jrayhawk> eh, it makes sense that the brain is limited by membrane remodelling capacity, which is mostly lipoprotein-driven
13:26 < jrayhawk> seth roberts famously increased his cognitive performance with large quantities of butter and then died of a massive heart attack
13:27 < kanzure> increasing cholesterol turnover via CYP46A1 (cholesterol 24-hydroxylase) causes better spatial learning and memory in mice https://journals.plos.org/plosone/article?id=10.1371%2Fjournal.pone.0068534
13:28 < kanzure> "The levels of N-Methyl-D-Aspartate receptor 1, phosphorylated-N-Methyl-D-Aspartate receptor 2A, postsynaptic density 95, synapsin-1 and synapthophysin were significantly increased in the hippocampus of the [humanized] CYP46A1 transgenic mice as compared to the controls. The levels of lanosterol in the brain of the CYP46A1 transgenic mice were significantly increased, consistent with a higher ...
13:28 < kanzure> ...synthesis of cholesterol. Our results are discussed in relation to the hypothesis that the flux in the mevalonate pathway in the brain is of importance in cognitive functions."
13:28 < L29Ah> MuaddibLLM: do saturated keto consumers die of CVD much more often?
13:28 < kanzure> "CYP46A1 catalyzes conversion of cholesterol into 24S-hydroxycholesterol, cholesterol to exit the brain."
13:28 < MuaddibLLM> checking recent meta-analyses on keto, saturated fat intake, and hard mortality endpoints...
13:29 < MuaddibLLM> Error: No valid text or tool use found in response. Retrying (1/3)...
13:29 < MuaddibLLM> leads to an error in the simulation, let's fix that.; ; The internal workings just failed, so I need to re-assess the information I have from the search results to formulate the final answer.; ; My findings so far:; 1.  **General Consensus vs. Nuance:** Traditional medical advice (and some studies like the Lancet paper [9] in search 2) suggests low-carb diets high in *animal* fat are associated with higher all-cause 
13:30 < MuaddibLLM> mortality compared to those with plant-based fat. However, newer meta-analyses and reviews (Search 2 [1], [6]) are often "inconclusive" or show "no increase" in CVD mortality for keto... full response: https://pasky.or.cz/muaddib/1bdd4edb0f3740dfb3bf4bb940987af4.txt
13:30 < jrayhawk> the other thing to consider is that there's tradeoffs between different fatty acids and cholesterol in building cellular membranes; if you feed neurons arachidonic acid and docosahexaenoic acid, how much do your cholesterol requirements go down?
13:31 < kanzure> "it is possible that enhanced ABCA1-mediated cholesterol efflux improves or preserves cognition"
13:32 < jrayhawk> and do LC-PUFAs-rich dendrite membranes allow long-term remodelling in ways that cholesterol-rich LC-SFA and LC-MUFA-based membranes require large amounts of substrate turnover?
13:33 < jrayhawk> LC-PUFA-rich membranes are inevitably going to be *much* more fluid
13:34 < jrayhawk> or less viscous, rather
13:40 < jrayhawk> larger lipoproteins having a harder time navigating the paracellular pathway would also inhibit atherosclerosis
13:41 < jrayhawk> probably with higher burdens on the rest of the vasculature and the liver
13:41 < jrayhawk> oxidative burdens
13:48 < kanzure> what about overexpression of DHA transporter Mfsd2a
13:50 < kanzure> "genetically induced bias toward DHA/AA incorporation perhaps via enhanced phospholipid acyltransferases"
13:50 < jrayhawk> re: masterjohn on statins: i am surprised he doesn't mention homeoviscocity altering dynamin's LDL invagination energy requirements, given that the big thing statins are doing is removing the capacity for the liver to produce cholesterol to mitigate SFA-driven membrane viscocity
13:51 -!- Guest88 [~Guest88@2a00:809:3b3:cd3e:e981:9f51:7873:51db] has quit [Quit: Client closed]
13:52 < jrayhawk> are there studies on increased LC-PUFA transport into dendrites using genetic copy number variation on the transporters?
13:53 < jrayhawk> the 500-1000mg per day limits do seem surprisingly small to me
13:54 < kanzure> "genetically induced bias toward DHA/AA incorporation perhaps via enhanced phospholipid acyltransferases"
13:54 < kanzure> "improving deficient DHA transport via MFSD2A in older brains could be a strategy to mitigate age-related cognitive decline" 
13:54 < kanzure> something something fat-1. i don't have a good world model for this.
13:58 < kanzure> "Overexpressing ACSL6 selectively increases DHA and AA internalization and incorporation into phospholipids, without similarly boosting saturated fats" https://pmc.ncbi.nlm.nih.gov/articles/PMC2846691/
14:02 < kanzure> also a recent study on FABP7 overexpression in hippocampus astrocytes "more spines and better synaptic function"... but this is a glial lipid chaperone not a transporter.
14:03 < kanzure> various benefits from fat1 overexpression.. but it is evidence of more LC-PUFA -> more dendrite spines -> better cognitive outcomes https://pmc.ncbi.nlm.nih.gov/articles/PMC2708766/
14:09 < hprmbridge> kanzure> and not transport specifically.
14:14 < jrayhawk> MuaddibLLM: by what mechanisms does lithium modify brain LC-PUFA metabolism?
14:14 < MuaddibLLM> Checking lithium's effect on the arachidonic acid cascade and PLA2 activity...
14:15 < MuaddibLLM> It specifically **downregulates cPLA2 (cytosolic phospholipase A2)** activity, effectively "braking" the release and turnover of Arachidonic Acid from membrane phospholipids (preventing the inflammatory COX/LOX cascade) while surprisingly increasing the synthesis of neuroprotective **17-OH-DHA**.
14:16  * fenn declares scrollback bankruptcy
14:18 < jrayhawk>  /set scrollback_lines 2048
14:36 < jrayhawk> all that fluidity has a cost in oxidative vulnerability, of course, so if you radically increase DHA transport, you might also need to experiment with Nrf2 and superoxide dismutase and such
14:39 < jrayhawk> s/radically/dramatically/ *Rodney Dangerfield necktie adjustment*
15:55 -!- darsie [~darsie@84-113-82-174.cable.dynamic.surfer.at] has quit [Ping timeout: 255 seconds]
16:06 < hprmbridge> kanzure> https://totalhealthoptimization.com/2025/11/24/nucleus-attempts-to-respond-but-accidentally-admits-its-own-guilt/
16:20 -!- flooded [~flooded@45.87.214.26] has quit [Remote host closed the connection]
16:42 < hprmbridge> kanzure> I need an LLM interface that simply picks the top latest frontier model on speed, benchmarks, and price within a certain range. Micromanaging daily choice is weird.
16:52 -!- stipa_ [~stipa@user/stipa] has joined #hplusroadmap
16:54 -!- stipa [~stipa@user/stipa] has quit [Ping timeout: 265 seconds]
16:54 -!- stipa_ is now known as stipa
17:00 < hprmbridge> kanzure> "Plastic surgery and all physicians have a medical license that lets them do anything within medicine that they feel comfortable to do. For example, a family medicine doctor is able to use their license to do brain surgery if they wish. An MD can legally attempt to perform any surgery they wish even with no prior experience or surgical training."
17:01 < L29Ah> MuaddibLLM: how to implant a human zygote? asking for a MD
17:01 < MuaddibLLM> Error: API error: RateLimitError("Error code: 429 - {'error': {'message': 'You exceeded your current quota, please check your plan and billing details. For more information on this error, read the docs: https://platform.openai.com/docs/guides/error-codes/api-errors.', 'type': 'insufficient_quota', 'param': None, 'code': 'insufficient_quota'}}")
17:02 < pasky> ohno
17:02 < pasky> hm maintaing credits across 3 different platforms is a bit more laborous
17:04 < pasky> MuaddibLLM: fixed
17:04 < MuaddibLLM> Back in business; looking up the manual for Zygote Intrafallopian Transfer vs. standard uterine payload delivery.
17:05 < pasky> hprmbridge: openrouter's autorouter could do it in theory, but at a single glance it didn't seem very transparent
17:05 < pasky> i mean kanzure ^
17:05 < MuaddibLLM> Technically, you don't—if your "MD" puts a Day 1 zygote directly into the uterus, it will likely fail to engraft due to endometrial asynchrony; you either incubate it to the blastocyst stage (Day 5) for standard transcervical catheter transfer, or you perform **Zygote Intrafallopian Transfer (ZIFT)**, which requires laparoscopic surgery to place it into the fallopian tube so it can develop en route 
17:05 < MuaddibLLM> (https://healthy.kaiserpermanente.org/health-wellness/health-encyclopedia/he.gamete-and-zygote-intrafallopian-transfer-gift-and-zift.hw202763).
18:30 -!- TMM [hp@amanda.tmm.cx] has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
18:30 -!- TMM [hp@amanda.tmm.cx] has joined #hplusroadmap
18:37 -!- flooded [~flooded@45.87.214.26] has joined #hplusroadmap
21:06 < hprmbridge> kanzure> "Rapid recovery: With minimal fine-tuning, excluded capabilities can be recovered extremely quickly - sometimes in under an hour. Latent risk: Signals linked to dual-use capabilities can persist in the pretrained representations and can be surfaced through simple methods like linear probing."
21:25 -!- flooded [~flooded@45.87.214.26] has quit [Remote host closed the connection]
21:26 -!- flooded [~flooded@45.87.214.26] has joined #hplusroadmap
21:54 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has quit [Read error: Connection reset by peer]
--- Log closed Tue Nov 25 00:00:48 2025