--- Log opened Fri Oct 03 00:00:57 2025
01:05 < fenn>  Volonaut Airbike - Real-life Speeder Bike "Forest Chase" (no chase, sadly) https://youtu.be/Q74qlF48qX0
01:07 < fenn> without music https://youtu.be/moSLCns3_rI
02:08 < hprmbridge> kanzure> Biomni, Kiin Bio, Kepler, Superbio, Camber, Tater Kosai, Lila, Periodic
02:09 < hprmbridge> kanzure> https://potato.ai/
02:30 -!- WizJin__ [~Wizzy@2402:a00:184:8354:28ae:ac51:4309:fcb4] has quit [Ping timeout: 252 seconds]
02:40 < hprmbridge> kanzure> terence tao vs chatgpt https://mathstodon.xyz/@tao/115306424727150237
02:40 < hprmbridge> kanzure> "potato" was not very creative. it took 20 minutes to make a pretty lame scientific plan.
03:03 < fenn> @Eli wow hellogoodlabs is waaaay cheaper than marek
03:04 < fenn> seems like they only do the blood donation thing at one location (downtown sf)
03:05 < fenn> kinda too bad because i would like to donate and get rid of some of my dirty old blood
03:25 < fenn> looks like there is a blood donation place in the same building as quest diagnostics nearby, and hellogoodlabs uses quest if you aren't donating
03:25 < fenn> maybe i will poke the blood donation people about joining
03:42 < fenn> "people who donated blood had an 88% lower risk of suffering a heart attack"
03:43 < fenn> @Eli doesn't that blow everything else out of the water, assuming it's causal
03:43 < hprmbridge> kanzure> how many times do they have to donate their blood to experience that effect?
03:43 < fenn> .t https://academic.oup.com/aje/article-abstract/148/5/445/76921
03:43 < saxo> Just a moment...
03:44 < fenn> "Donation of Blood Is Associated with Reduced Risk of Myocardial Infarction: The Kuopio Ischaemic Heart Disease Risk Factor Study" it's got finns so you know it's good
03:46 < fenn> ". In a Cox proportional hazards model adjusting for age, examination years and all other predictive coronary disease risk factors, blood donors had a 88% reduced risk" N=2,862 (153 donors, 1 heart attack vs 2,529 non-donors and 316 heart attacks)
03:49 < fenn> it's a very wide hazard ratio confidence interval 0.02–0.86 (they used 0.12 as the hazard ratio)
03:53 < fenn> they should also compare involuntary blood donors
03:59 < fenn> "risk of coronary disease in women is increased by simple hysterectomy without oophorectomy, which suggests that menstrual bleeding might play an important role in the risk reduction."
04:01 < fenn> also should compare to plasma donors
04:01 < fenn> MuaddibLLM: make it so, find me some plasma donor stats
04:03 < MuaddibLLM> hunting for plasma donor stats, because apparently we need to compare every possible way of getting blood out of people...
04:03 < fenn> it's not the same stuff!
04:04 < MuaddibLLM> Plasma donation lowers cholesterol (up to 47 mg/dL total, 34 mg/dL LDL in high-baseline donors at 2-4 day intervals) https://pmc.ncbi.nlm.nih.gov/articles/PMC3813983/ and BP (~20/9 mmHg in Stage 2 hypertensive donors) https://pmc.ncbi.nlm.nih.gov/articles/PMC4302974/ - but unlike blood donation's 88% MI risk reduction, no direct CVD outcome studies exist yet (mechanism likely differs: iron depletion vs lipid/protein 
04:04 < MuaddibLLM> removal).
04:07 < fenn> so this all seems very straightforward and obvious, why is it not common knowledge and recommended preventative medical advice
04:08 < fenn> are they afraid of looking like barbaric blood-letters?
04:08 < MuaddibLLM> [gpt-5] Because the data are mostly observational and surrogate-only (lipids/BP, not hard outcomes) with healthy-donor selection bias, likely require impractically frequent draws with IgG/albumin tradeoffs, lack any RCTs showing fewer MIs/strokes, and run into ethics/liability/logistics of “prescribing” donation (plus industry data silos) - so guidelines won’t endorse it yet.
04:09 < fenn> i mean the hypothesis that reduction in iron stores prevents heart attacks by reducing lipid oxidation
04:09 < fenn> i don't see any ethics problems with prescribing donation, the opposite in fact
04:10 < fenn> you don't have to exercise either, if "prescribed"
04:10 < fenn> has anyone ever had a doctor tell them to donate blood, for its health benefits?
04:10 < fenn> or had anyone tell them that, at all
04:11 < MuaddibLLM> [gpt-5] Because the big RCT killed it - FeAST (2007) found phlebotomy in 1277 PAD patients did nothing (HR 0.85, p=0.17) despite lowering ferritin 122→80 ng/mL, and recent Mendelian randomization studies show HIGHER iron status is actually protective against CAD https://pubmed.ncbi.nlm.nih.gov/28684612/ - that 88% blood donor benefit was likely pure healthy-donor selection bias, the oxidation mechanism didn't pan out in 
04:11 < MuaddibLLM> mechanistic studies.
04:12 < fenn> mendelian randomization?
04:14 < fenn> © 2017 American Heart Association, Inc. 
04:15 < fenn> MuaddibLLM: "protective effect of higher iron status on CAD risk (iron odds ratio, 0.94 per SD" is pretty wimpy effect right
04:16 < MuaddibLLM> Yep - 0.94 per SD is ~6% lower CAD odds per SD of genetically higher iron, so modest per person but still meaningful over 2–3 SD (~12–17%) and, as a lifelong causal estimate, it makes that 88% donor effect look like selection bias rather than physiology.
04:16 < fenn> i wouldn't want to be 3 SD high in iron
04:18 < fenn> MuaddibLLM: i'm having trouble understanding what genetics has to do with this experiment design
04:18 < MuaddibLLM> Mendelian randomization uses genetic variants that nudge iron status up/down (e.g., HFE, TMPRSS6, transferrin-related) as quasi-random lifelong “assignments,” so by seeing how CAD risk shifts per genetically-predicted iron SD you get a causal estimate with less confounding/reverse causation than donor vs non-donor comparisons - caveats are pleiotropy and possible nonlinearity.
04:19 < fenn> MuaddibLLM: why not pick an unrelated gene or even non-coding region
04:20 < fenn> or no gene at all and just use the last digit of SSN..
04:20 < MuaddibLLM> In MR you only pick variants that strongly shift the exposure (here iron) - random/unrelated SNPs give weak‑instrument bias and violate the exclusion restriction, while non‑coding is totally fine if it robustly alters iron (eQTL/regulatory hits); standard practice is clumped genome‑wide significant iron SNPs plus pleiotropy guards (IVW, MR‑Egger, weighted median, MR‑PRESSO).
04:22 < fenn> i don't get it. a genotype that increases iron stores is not the same as modifying your diet or blood
04:23 < MuaddibLLM> [gpt-5] Right - MR doesn’t equate genotype with diet/phlebotomy; it uses genes as quasi-random lifelong nudges to iron to estimate the effect of iron itself on CAD (a LATE for small, chronic shifts), which should match per-unit iron changes if the biology is roughly linear and path-independent, but can break for acute/large or pathway-specific interventions like bloodletting or chelation.
04:24 < fenn> it's not random because you'd expect higher iron variants to be correlated with higher levels of active protection from that extra iron
04:25 < fenn> the alleles are not independent
04:26 < fenn> no sir, i don't like it
04:26 < fenn> let's do the stupid obvious thing that seems to work and benefits everybody
04:30 < hprmbridge> kanzure> no you can't do that!! havent you forgotten to ask an ethics council???
04:33 < hprmbridge> kanzure> I've been thinking about an award ceremony focused on granting awards to the top innovative projects and opportunities that have been specifically canceled by regulation or red tape. Currently there is no way for those kinds of projects to get high visibility and people kind of shrug and move on and say, oh well. The negative consequences of regulations needs to be more publicly visible. There
04:33 < hprmbridge> kanzure> should be a dashboard. There should be a way to query and quickly see, here's the estimated impact per year of each regulation for delaying the development of these technologies and the distribution of these technologies.
04:36 < fenn> sinai award? https://www.tabletmag.com/feature/sinai-awards-2025
04:37 < fenn> hmm nevermind
04:37 < fenn> who would you nominate for the tape-ducker award
04:38 < fenn> oh the award is FOR getting canceled
04:38 < fenn> i was thinking people who went and did stuff anyway
04:39 < hprmbridge> kanzure> well there should be a few categories
04:39 < hprmbridge> kanzure> including things that were shelved early just because of the perception of excess regulation
04:39 < jrayhawk> much like pushing low-SFA diets on everyone because 10% of the population has a lipoprotein processing disorder, pushing phlebotomy on everyone because 15% of the population has some degree of hemochromatosis isn't exactly wise
04:40 < hprmbridge> kanzure> can you test to see
04:40  * fenn mumbles about the broken window fallacy
04:40 < fenn> how would you discover all the things that were not done
04:40 < fenn> it's uncountable
04:42 < hprmbridge> kanzure> it doesn't matter. you just need some visibility of the problem.
04:43 < fenn> huh. "People of Northern European descent have a particularly high incidence of hemochromatosis type 1, with about 1:8 people being carriers of the principal genetic variant, the C282Y mutation on the HFE gene, and 0.5% of the population having the condition."
04:43 < fenn> i suppose the heterozygotes have impaired function
04:44 < fenn> "It is thought that this protein functions to regulate circulating iron uptake by regulating the interaction of the transferrin receptor with transferrin."
04:47 < jrayhawk> more whisper churn https://signal.org/blog/spqr/
04:50 < jrayhawk> it's a black plague adaptation; it hides iron away from circulating infectious bacteria
04:51 -!- justanot1 [~justanoth@gateway/tor-sasl/justanotheruser] has quit [Ping timeout: 272 seconds]
04:51 < fenn> yes that was the main reason for blood letting
04:51 < fenn> protein status is much better now
04:51 < fenn> more antibodies, less risk of infection
05:04 -!- justanotheruser [~justanoth@gateway/tor-sasl/justanotheruser] has joined #hplusroadmap
05:31 -!- TMM [hp@amanda.tmm.cx] has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
05:31 -!- TMM [hp@amanda.tmm.cx] has joined #hplusroadmap
05:48 -!- justanotheruser [~justanoth@gateway/tor-sasl/justanotheruser] has quit [Ping timeout: 272 seconds]
05:59 -!- hellleshin [~talinck@172-2-128-195.lightspeed.cntmoh.sbcglobal.net] has quit [Quit: Connection reset by bear]
06:00 -!- helleshin [~talinck@172-2-128-195.lightspeed.cntmoh.sbcglobal.net] has joined #hplusroadmap
06:07 -!- helleshin [~talinck@172-2-128-195.lightspeed.cntmoh.sbcglobal.net] has quit [Read error: Connection reset by peer]
06:09 -!- helleshin [~talinck@172-2-128-195.lightspeed.cntmoh.sbcglobal.net] has joined #hplusroadmap
06:18 -!- srk [~sorki@user/srk] has quit [Ping timeout: 264 seconds]
06:18 -!- srk [~sorki@user/srk] has joined #hplusroadmap
06:18 -!- Malvolio [~Malvolio@idlerpg/player/Malvolio] has quit [Ping timeout: 264 seconds]
06:18 -!- superkuh [~superkuh@user/superkuh] has quit [Ping timeout: 264 seconds]
06:35 -!- superkuh [~superkuh@user/superkuh] has joined #hplusroadmap
07:23 -!- Malvolio [~Malvolio@idlerpg/player/Malvolio] has joined #hplusroadmap
07:49 -!- flooded [~flooded@149.102.226.202] has joined #hplusroadmap
08:13 -!- catalase [~catalase@user/catalase] has quit [Killed (NickServ (GHOST command used by catalase0))]
08:13 -!- catalase0 [~catalase@user/catalase] has joined #hplusroadmap
08:27 < kanzure> horror story: i spent twenty minutes chatting with a voice agent and only at the end realized its cutoff date was 2023. it was *bad*.
08:28 < kanzure> it was able to speak english but not able to actually say anything with the words it used;  and it had no conception that its paragraphs were devoid of knowledge.
08:46 < pasky> what voice agent was it?
08:47 < hprmbridge> kanzure> chatgpt
08:48 < pasky> oh interesting. i used it a few times during driving, often to chat about a particular website i had no time to read or to have some kind of "partner" for dumping my thoughts.
08:48 < pasky> it's right on the edge of usability. sometimes over the edge, sometimes not.
08:48 < hprmbridge> kanzure> it was only able to rephrase my question and emphasize how right i was, rather than telling me any information. like in surgical technique or wasmtime topics.
08:49 < pasky> also the app has many bugs around the voice mode - sometimes part of the chat history is not visible in the app, you can't share the conversation, ...
08:49 < pasky> hm, interesting. i wasn't actually ever asking it for learned knowledge i think, though.
08:49 < pasky> my experience is if you fill its context with the right things, it can be useful. it can browse the web.
08:50 < hprmbridge> kanzure> well if it can, it did not do that for me!
08:50 < pasky> you have to ask it explicitly :)
08:51 < pasky> basically it's like if i had a high-school-educated secretary on the phone\
08:51 < pasky> (eager for promotion but not smart)
08:52 < pasky> "look up something for me and read / summarize it" is the best you can hope for but sometimes it's exactly what you need
09:00 < pasky> i did expect to see a major upgrade of the voice mode model after gpt-realtime release, though... maybe soon
10:17 < kanzure> this keeps going through my head every day all day https://diyhpl.us/~bryan/irc/neoslop.png except for claude
10:18 -!- justanotheruser [~justanoth@gateway/tor-sasl/justanotheruser] has joined #hplusroadmap
10:59 < kanzure> "In 2019, Ian Clarke began work on a successor to the original Freenet, which was internally known as “Locutus.” This project, a redesign from the ground up, incorporated lessons learned from the original Freenet’s development and operation, and adapted to today’s challenges. In March 2023, the original version of Freenet was separated into its own project (Hyphanet), and what was ...
11:00 < kanzure> ...known as “Locutus” was officially renamed to “Freenet.”"
11:01 < kanzure> https://www.hyphanet.org/ https://www.hyphanet.org/pages/about.html
11:01 < kanzure> okay.... so what's modern freenet then?
11:26 < hprmbridge> Eli> Interesting. That would be a big effect size. I would be curious about controlling for pharma meds. You can't really give blood if you're on certain medicine. Also, big CI. I've read that for hemochromatosis, blood letting is a treatment. I've been told that hemochromatosis is a risk factor for Alzheimer's. So, the stuff that doctors were doing hundreds of years ago actually had a benefit for some
11:26 < hprmbridge> Eli> people ...
11:44 < hprmbridge> poppingtonic> https://tangu-kale-labs-web.fly.dev/
12:13 < hprmbridge> kanzure> your site?
12:14 < kanzure> freenode's manuscripts are all riddled with very weird grammar errors. maybe i'm having a stroke because the wikipedia page is also full of weirdly written english.
12:17 < kanzure> who uses this?
12:28 < kanzure> "It is important to note that the maintainers of the original Freenet did not agree with the decision to rearchitect and rebrand." no kidding?
13:02 -!- ike8 [e8f913dbdf@irc.cheogram.com] has quit [Ping timeout: 252 seconds]
13:14 -!- ike8 [e8f913dbdf@irc.cheogram.com] has joined #hplusroadmap
14:21 < hprmbridge> poppingtonic> It is maintained by a small lab I collaborate with. My friend Ben Mainye built it
14:52 -!- darsie [~darsie@84-113-82-174.cable.dynamic.surfer.at] has quit [Ping timeout: 256 seconds]
16:53 -!- stipa_ [~stipa@user/stipa] has joined #hplusroadmap
16:56 -!- stipa [~stipa@user/stipa] has quit [Ping timeout: 240 seconds]
16:56 -!- stipa_ is now known as stipa
17:02 < hprmbridge> kanzure> https://proteinbase.com/proteins
17:44 < hprmbridge> kanzure> "Left-handed DNA for efficient highly multiplexed imaging at single-protein resolution" https://www.nature.com/articles/s41467-025-64228-x (2025) DNA-PAINT but with 12 targets
18:17 < L29Ah> switched from sprattus sprattus to clupea harengus membras, wtf it is a lot harder to butcher per piece
19:25 -!- catalase [~catalase@user/catalase] has joined #hplusroadmap
19:28 -!- catalase0 [~catalase@user/catalase] has quit [Ping timeout: 264 seconds]
20:48 -!- TMM [hp@amanda.tmm.cx] has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
20:48 -!- TMM [hp@amanda.tmm.cx] has joined #hplusroadmap
21:31 -!- justanotheruser [~justanoth@gateway/tor-sasl/justanotheruser] has quit [Remote host closed the connection]
21:31 -!- justanotheruser [~justanoth@gateway/tor-sasl/justanotheruser] has joined #hplusroadmap
23:17 -!- darsie [~darsie@84-113-82-174.cable.dynamic.surfer.at] has joined #hplusroadmap
--- Log closed Sat Oct 04 00:00:58 2025