--- Log opened Tue Jan 06 00:00:28 2026
03:55 -!- darsie [~darsie@84-113-82-174.cable.dynamic.surfer.at] has joined #hplusroadmap
06:24 < jrayhawk> https://scholar.google.com/citations?user=NCQ3E5sAAAAJ did a lot of directed longevity evolution experiments
06:26 < kanzure> yes he has the methuselah drosophila
06:27 < jrayhawk> he had a funny presentation at an AHS a while back https://www.youtube.com/watch?v=GkUEe_UGP0I
06:29 < jrayhawk> which was arguing that shallow adaptation to a local minimum will bias towards salvaging earlier productive years and away from later ones, so it may sometimes be optimal to switch to an older input distributions halfway through your reproductive years
06:30 < jrayhawk> er, s/ productive/ reproductive/
06:32 <+gnusha> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=345a1cd7 Bryan Bishop: transcript: living longer with michael rose >> https://diyhpl.us/wiki/transcripts/living-longer-michael-rose/
06:43 < kanzure> https://computeprices.com/
06:52 < kanzure> well let's see what michael rose said in his emails https://diyhpl.us/~bryan/irc/chatgpt/grg/switching-diets-early.html
07:10 <+gnusha> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=214bb557 Joe Rayhawk: nutrition/framing: clean up and extend opioid discussion >> https://diyhpl.us/wiki/nutrition/gv/receptor-opioid-current.gv
07:12 <+gnusha> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=7dc48501 Joe Rayhawk: nutrition/framing: typo >> https://diyhpl.us/wiki/nutrition/framing/
07:17 <+gnusha> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=99ddf610 Joe Rayhawk: nutrition/framing: fix ordered sublist, hopefully >> https://diyhpl.us/wiki/nutrition/framing/
07:19 <+gnusha> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=d2d910e5 Joe Rayhawk: nutrition/framing: make tomatine link go directly to glycoalkaloid cytotoxicity graph >> https://diyhpl.us/wiki/nutrition/framing/
07:24 -!- flooded [~flooded@149.102.226.195] has joined #hplusroadmap
08:15 <+gnusha> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=9b343de0 Bryan Bishop: add information about DNA synthesis chemistry reagents and precursors >> https://diyhpl.us/wiki/dna/synthesis/notes/
08:39 < kanzure> looking over all these GRG results, i notice that either the LLMs are very one shotted or the GRG members are one shotted on individual supplements or drugs as cure-alls
08:39 < kanzure> no ability to sustain multiple months of thinking about feedback loops, recursion, causality
08:40 < kanzure> poly-causal reasoning is present but not in any meaningful way
08:41 < kanzure> also, people don't seem to be careful enough about which selection effects they are either exposing themselves to, or their thinking to (e.g. when looking at supercentenarians you are actually looking through several fun subtle sets of selection effects)
08:43 < kanzure> MuaddibLLM: in a single sentence, characterize andropause for a highly technical audience
08:43 < MuaddibLLM> kanzure: Andropause is the gradual, non-discrete decline in Leydig cell testosterone synthesis and hypothalamic-pituitary-gonadal axis sensitivity beginning around age 30-40, characterized by falling total/free testosterone (~1-2%/year), rising SHBG, blunted LH pulsatility, and downstream effects on muscle mass, bone density, erythropoiesis, libido, and cognitive function - distinct from menopause in that it lacks a 
08:43 < MuaddibLLM> defined cessation event and exhibits high inter-individual variability in symptom threshold.
08:43 < MuaddibLLM> (this message used 0 tool calls, 8544 in / 257 out tokens, and cost $0.0491)
08:43 < kanzure> that message cost $0.05 ?!
08:44 < L29Ah> that's his 5c
08:45 < L29Ah> MuaddibLLM: should andropause TRT also include estrogen?
08:45 < MuaddibLLM> Checking whether TRT protocols should include estrogen or if aromatization handles it...
08:46 < MuaddibLLM> L29Ah: No - exogenous testosterone aromatizes to estradiol naturally, so you get estrogen "for free"; the actual clinical concern is usually the *opposite* direction (excess E2 causing gynecomastia/water retention), hence AI prescriptions in some protocols - but you do need to preserve *some* estradiol because it's essential for male bone mineralization, libido, and cardiovascular function (per JCI 2024: "T's conversion to 
08:46 < MuaddibLLM> E2 is fundamental to bone and vascular health"), so aggressive aromatase inhibition is counterproductive.
08:46 < MuaddibLLM> (this message used 0 tool calls, 9454 in / 459 out tokens, and cost $0.0587)
08:46 < pasky> hm that's weird, right?
08:47 <+gnusha> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=66feca5c Bryan Bishop: longevity, before i forget >> https://diyhpl.us/wiki/longevity/
08:49 < pasky> why is it 8544 tokens :/
08:51 <+gnusha> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=c5efc4bf Bryan Bishop: some education links >> https://diyhpl.us/wiki/education/
08:51 <+gnusha> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=59164578 Bryan Bishop: some links on alternative education >> https://diyhpl.us/wiki/education/
09:06 < pasky> just made a change that might save me 50% of cost :)
09:28 < kanzure> pasky: you gotta use the --burn-less-tokens flag
09:36 < kanzure> https://diyhpl.us/~bryan/irc/chatgpt/grg/feedback-loops.html
09:37 < kanzure> for autoimmune damaged individuals, why not put them into into positive‑pressure sterile isolation bubble units?
09:37 < kanzure> oh and ablate their bone marrow
09:44 < pasky> oh i gotta tell ampcode to implement a --burn-less-tokens flag :)
09:47 < kanzure> what about: refactor all human proteins such that they are able to survive the most likely ribosomal errors in old age. a similar idea was mentioned in genetic-modifications.mdwn but it was primarily about preserving functionality in general, not looking at likely translation errors, amino acid incorporation errors, or folding errors (including folding chaperone or cofactor problems) in old ...
09:47 < kanzure> ...age.
09:48 < kanzure> "RAID-1 genomes: duplicate key proteostasis genes (e.g., 10x [folding] chaperones)." ok
09:51 < kanzure> MuaddibLLM: search the research literature for methods for enhancement of the endoplasmic reticulum with regards to protein degradation of unfolded or misfolded proteins, BiP or ERdj5, unfolded-protein response (UPR), etc. In particular look for longevity or lifespan outcomes from transgenics or modifications here.
09:51 < MuaddibLLM> Digging through ER proteostasis literature for lifespan extension hits...
09:52 < MuaddibLLM> ScienceDirect being annoying, hitting PMC for the good stuff...
09:52 < MuaddibLLM> kanzure: Strong hits - (1) neuronal XBP1s overexpression extends C. elegans lifespan and restores cognitive function in aged mice by activating UPR in distal tissues (Cell 2013, EMBO J 2022: https://pmc.ncbi.nlm.nih.gov/articles/PMC9670206/); (2) SKN-1A/Nrf1 overexpression (ER-tethered TF that upregulates proteasome subunits in response to misfolded proteins) extends C. elegans lifespan ~20% and is necessary for healthy 
09:52 < MuaddibLLM> aging (eLife 2019: https://elifesciences.org/articles/44425); (3) EDEM overexpression (ERAD-enhancing mannosidase-like proteins) protects against chronic ER proteinopathy; (4)... full response: https://m.or.cz/z7TJwO6H.txt
09:52 < MuaddibLLM> (this message used 0 tool calls, 40299 in / 1569 out tokens, and cost $0.2407)
09:53 < kanzure> pasky: how did that use 0 tool calls if it has good links?
10:05 <+gnusha> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=93c3e921 Bryan Bishop: increase nots on ribosomal hyperaccuracy and endoplasmic reticulum improvements >> https://diyhpl.us/wiki/genetic-modifications/
10:05 < kanzure> notes :|
10:07 <+gnusha> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=c289d8c9 Bryan Bishop: also more chaperone copies >> https://diyhpl.us/wiki/genetic-modifications/
10:29 -!- WizJin_ [~Wizzy@43.249.228.126] has joined #hplusroadmap
10:31 < kanzure> "gene-length-dependent transcription decline"
10:31 < kanzure> "extremely long synaptic genes at chromosomal fragile sites and their vulnerability to postmitotic DNA damage."
10:32 -!- WizJin__ [~Wizzy@150.129.166.255] has quit [Ping timeout: 252 seconds]
10:33 < kanzure> "Time is ticking faster for long genes in aging" https://pmc.ncbi.nlm.nih.gov/articles/PMC11003850/ (2025)
10:33 < kanzure> oh, 2024. well whatever.
10:50 -!- WizJin__ [~Wizzy@43.249.228.126] has joined #hplusroadmap
10:54 -!- WizJin_ [~Wizzy@43.249.228.126] has quit [Ping timeout: 264 seconds]
10:55 < kanzure> "Methylation of ribosomal RNA by NSUN5 is a conserved mechanism modulating organismal lifespan" https://www.nature.com/articles/ncomms7158 (2015)
10:55 < kanzure> the great wise slop machine says: "When NSUN5 activity is reduced, ribosomes lack one specific m⁵C mark in large-subunit rRNA. Those ribosomes translate most housekeeping mRNAs slightly worse but translate stress-response and redox-maintenance mRNAs disproportionately better. The result is a proteome biased toward maintenance rather than growth."
11:01 <+gnusha> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=6a6e0657 Bryan Bishop: rRNA bias towards maintenance, other nucleolus things >> https://diyhpl.us/wiki/genetic-modifications/
11:08 < kanzure> what about: take supercentenarian cell biopsy and use these as inputs in directed evolution projects, to select for cells that are particularly resistant to aging. for example, they could be cultured and then used in animal longevity trials, followed by further downstream selection.
11:18 -!- etc-vi [~etc-vi@user/meow/girlchunks] has joined #hplusroadmap
11:18 < pasky> kanzure: the tool counter is broken and it's not obvious to me why and claude is telling me it's not broken and it's a total mystery
11:21 < L29Ah> pathetic python vibe code
11:25 < pasky> indeed indeed
12:20 < hprmbridge> nmz787> https://nikomc.com/posts/heteropolymers.html
12:20 < hprmbridge> nmz787> """New Blog: A new paper shows that enzymes can be made by mixing just four molecules together, none of which are amino acids. The four molecules randomly link together to form long polymer chains, some of which catalyze chemical reactions."""
12:20 < hprmbridge> nmz787>
12:20 < hprmbridge> nmz787> """
12:20 < hprmbridge> nmz787> Methyl methacrylate (MMA), a hydrophobic molecule.
12:20 < hprmbridge> nmz787> 2-ethylhexyl methacrylate (EHMA), an even more hydrophobic molecule.
12:20 < hprmbridge> nmz787> Oligo(ethylene glycol) methyl ether methacrylate (OEGMA), a hydrophilic molecule.
12:20 < hprmbridge> nmz787> 3-sulfopropyl methacrylate potassium salt (SPMA), which mimics aspartate as an electron donor; the active site surrogate."""
12:20 < hprmbridge> nmz787> https://www.nature.com/articles/s41586-025-09860-9
12:25 < kanzure> is this the paper about randomization of polymers to make enzymes
12:25 < kanzure> yes ok
12:25 < kanzure> similar concept to aptamers, DNAzymes or ribozymes except using a different chemistry
12:27 < kanzure> or drmeister's bis amino acids
12:27 < kanzure> instead of alpha amino acids
12:28 < kanzure> the advantage of randomized nucleic acid polymers is that you can turn around and sequence them and get reusable genetic information
12:31 < L29Ah> you can sequence this shit too if you're so inclined
12:32 < L29Ah> but synthesizing a bunch of random shit is much much cheaper than a specific sequence
12:32 < L29Ah> so maybe you'd want to productionalize shit regardless if you know the sequence
12:45 < kanzure> "My favourite statistical "proof" about the limits of longevity was made by L Hayflick in (How & Why We Age, 1994) in which he estimated that the 120-year barrier wouldn't be reached for ..... ten-thousand years. Jeanne Calment reached 120 next year. Hayflick also thought the maximum human LS was 115, despite JC already being 119 at the time."
12:49 -!- TMM [hp@amanda.tmm.cx] has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
12:49 -!- TMM [hp@amanda.tmm.cx] has joined #hplusroadmap
13:23 < hprmbridge> nmz787> Presumably JC isn't Jesus Christ?
13:31 <+gnusha> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=757471e2 Bryan Bishop: add transcripts of various longevity lectures >> https://diyhpl.us/wiki/transcripts/gerontology-research-group/young-plasma-results-2024/
13:35 < kanzure> jeanne calment was not jesus christ, correct
13:50 <+gnusha> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=01726c6a Bryan Bishop: fix aubrey de grey transcript from 2009 >> https://diyhpl.us/wiki/transcripts/longevity-summit-dublin/enhancing-dna-repair-to-reverse-senescence/
14:20 < kanzure> is there any casual explanation for why minor genetic modifications in mice extends lifespan so often? i mean 10-20% is kind of a large effect. but why that effect size in particular? is this a consequence of study design and low population size in the study? is it because other effects dominate after 20%?
14:21 < L29Ah> mice are r-strategists
14:21 < L29Ah> you won't notice them reducing their breeding rate but that's what they optimize for, not longevity
14:22 < kanzure> AFAIK aubrey is still seeking funding for his multi-genetic-intervention robust mouse rejuvenation project. he is asking for $5m i think?
14:22 < kanzure> and a few thusand mice
14:22 < kanzure> thousand...
14:25 < kanzure> TODO: fix endosome ESCRT-mediated repair failure; fix impaired endosomal maturation; fix tubulin hyperacetylation; detyrosination of microtubules; lamin aggregation; intermediate filament phosphorylation errors; improve localization signalers or chaperones or whatever; reduce lipid peroxidation; reduce centrasomal fragmentation. don't know why this was in an open document but i'm closing it ...
14:25 < kanzure> ...now.
14:27 < kanzure> i added a RAG LLM pass to clean up the context using a large language model. And no matter what I prompt, it completely rewrites the context instead of merely removing email formatting.
15:31 < hprmbridge> Eli> I was kind of thinking about the recent opus update. I'm wondering if we might be headed into a world where AI is going to be able to trivially create software. Like, is one man going to be able to vibe code a photoshop or excel program into existence if they just do enough prompting? If we keep going down this path of superintelligence, what does the software industry look like where the MRR of
15:31 < hprmbridge> Eli> all these existing companies falls to $0? What new companies end up dominating?
15:43 < pasky> one annoying thing about vibecoding universal software is that there is a lot of cornercases (around real world conditions, unexpected [by you] needs, security issues..) ... even if you have vibecoding ASI at your hand, it's unlikely to see all these in advance, and then your software breaks and your ASI needs to fix it and it's just annoying and more convenient to get a premade piece of software 
15:43 < pasky> where someone met and fixed these edge cases before you
15:43 < pasky> (perfect example is your os - i think it'll be a very long time before we replace os with something fully ai generated)
15:43 < pasky> *on the other hand* the huge power of vibecoding is you don't need to vibecode universal software but something that just ad hoc solves a task you need
15:44 < pasky> and the less universal software is, the less edge cases there are to cover
15:44 < pasky> so you won't be vibecoding your photoshop, you'll just ask your AI to replace a background / make it sepia
15:45 < pasky> why create your excel when you can ask the AI to create a business model for you? (it'll do it in pandas and visualize the way you want)
16:04 < L29Ah> Eli: hey chatgpt code me a program that figures out immortality, quantum gravity and everything
16:22 < pasky> hmm looks like my gadget to locally mirror my twitter history so that I can grep my likes is working pretty well now
16:23 -!- stipa_ [~stipa@user/stipa] has joined #hplusroadmap
16:26 -!- stipa [~stipa@user/stipa] has quit [Ping timeout: 250 seconds]
16:26 -!- stipa_ is now known as stipa
16:48 < hprmbridge> Eli> Interesting point. Just have ai do all of it on the fly. Your OS could be barebones and everything runs through ai in the cloud
16:49 < hprmbridge> Eli> Then you get your BCI and think tasks into existence
16:52 -!- darsie [~darsie@84-113-82-174.cable.dynamic.surfer.at] has quit [Ping timeout: 240 seconds]
17:29 < kanzure> "Somatic mutation rates scale with lifespan across mammals" https://pmc.ncbi.nlm.nih.gov/articles/PMC9021023/ (2022)
20:09 -!- dustinm [~dustinm@static.38.6.217.95.clients.your-server.de] has quit [Ping timeout: 264 seconds]
20:22 -!- dustinm [~dustinm@static.38.6.217.95.clients.your-server.de] has joined #hplusroadmap
23:24 <+gnusha> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=5f0f30f4 fenn: elastin repair issues powerpoint slide >> https://diyhpl.us/wiki/users/fenn/longevity elastin interventions and pitfalls by lolo nano-banana jIzfk.32.33.png
23:27 <+gnusha> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=82fd86cf fenn: see if underscores makes ikiwiki happy >> https://diyhpl.us/wiki/users/fenn/longevity_elastin_interventions_and_pitfalls_by_lolo_nano-banana_jIzfk.32.33.png
23:29 <+gnusha> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=b953c705 fenn: filename changed; links need underscores >> https://diyhpl.us/wiki/longevity/
23:30  * fenn shrugs
23:32 < fenn> "OS could be barebones and everything runs through ai in the cloud" dystopia
23:32 < fenn> you will own nothing and be happy
23:32 < fenn> or else
23:43 -!- etc-vi [~etc-vi@user/meow/girlchunks] has quit [Quit: nya]
--- Log closed Wed Jan 07 00:00:29 2026