--- Log opened Wed Aug 13 00:00:09 2025
00:35 < fenn> can japan pull off a world record lowest birth rate in 2026? https://en.wikipedia.org/wiki/Fire_Horse#Superstition
00:45 -!- TMM [hp@amanda.tmm.cx] has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
00:46 -!- TMM [hp@amanda.tmm.cx] has joined #hplusroadmap
01:37 -!- justanotheruser [~justanoth@gateway/tor-sasl/justanotheruser] has quit [Ping timeout: 240 seconds]
04:08 -!- stipa [~stipa@user/stipa] has quit [Quit: WeeChat 4.6.3]
04:08 -!- stipa [~stipa@user/stipa] has joined #hplusroadmap
04:30 < hprmbridge> kanzure> "Two-photon 3D printing of functional microstructures inside living cells" https://arxiv.org/abs/2506.13232
04:31 < hprmbridge> kanzure> "A biocompatible photoresist is injected into live cells and selectively polymerized with a femtosecond laser. The unpolymerized photoresist is dissolved naturally within the cytoplasm, leaving behind stable intracellular structures with submicron resolution within live cells. We printed various shapes, including a 10 micron elephant, barcodes for cell tracking, diffraction gratings for remote
04:31 < hprmbridge> kanzure> readout, and microlasers."
04:33 < hprmbridge> kanzure> there was previously a paper posted in here that was doing 3d printing inside the human body thru skin without surgery; here the idea is further miniaturized.
04:34 < hprmbridge> kanzure> can this be used as an alternative to natural cytoskeletal cell shape determination?
04:38 < hprmbridge> kanzure> "One of the applications we explored is barcoding, which involves writing a specific code to each cell for the purpose of identification and long-term tracking of the cell. We have designed a 3D graphical barcode composed of 4 stacked 4x4 grids of cylinders (Fig. 4a, Supplementary Fig. 5c, and Supplementary Video 3). Each of the 64 spaces can be occupied or not. In the middle two layers, at least
04:38 < hprmbridge> kanzure> one cylinder should be present so that the structure does not fall apart. One cylinder on the top corner is also always present and is star-shaped so that the orientation of the structure can be determined. Therefore, we can encode 61 bits of information, giving a total of (2x10^18) unique barcodes. This number far exceeds the number of cells in the human body. For real-life applications, much
04:38 < hprmbridge> kanzure> smaller barcodes could be used. Compared to most other cell barcoding techniques [21, 26], our method uniquely allows for encoding predefined information into each cell, rather than relying on random barcodes."
04:40 < hprmbridge> kanzure> "Instead of using photoresists only slightly soluble in water, it is also possible to inject a water-soluble hydrogel-based photoresist into the cells. In this case, the photoresist gets distributed throughout the cell, enabling printing anywhere within, thus making it possible to isolate a certain part of the cell inside a newly made compartment, for example, to study the individual functioning
04:40 < hprmbridge> kanzure> of certain organelles or to investigate signaling pathways by physically blocking some of them. In this approach, instead of inactivating or physically destroying organelles to study their regeneration or biogenesis [29, 30], it would be possible to fix them or entire selected cells by photo-polymerization."
04:43 < hprmbridge> kanzure> why wouldn't you simply microinject your tiny printed micro-elephant instead of doing photoresist polymerization printing inside of a cell?
04:44 < hprmbridge> kanzure> you are already microinjecting this starting material, after all.
04:47 < hprmbridge> kanzure> maybe some fun applications for physarum polycephalum?
04:53 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has joined #hplusroadmap
05:08 < hprmbridge> Eli> Based on the link, they’re just giving mice rapa and combining it with other stuff. I don’t think we are going to get average lifespan much beyond 100 doing this. It would be hugely beneficial for society to  reduce years spent in disability. So, if it works it would be good. But these pathways being affected won’t make us live forever.
05:08 < hprmbridge> Eli>
05:08 < hprmbridge> Eli> Sinclair is claiming rejuvenation, but he has made claims in the past and they didn’t turn out well.
05:09 < fenn> OSK is a plausible root cause mechanism for programmed aging
05:10 < hprmbridge> kanzure> silence, luddite! we shall not stop until mice live for millions of years.
05:10 < hprmbridge> kanzure> big cheese compels us to this vision
05:14 < hprmbridge> Eli> Yeah, I think Sinclair is claiming that’s what they’re going to do. It would be nice if they can actually prove it. He got in trouble with resveratrol and NMN claims.
05:15 < fenn> it would be nice if other people besides sinclair were studying it
05:15 < fenn> 99% of "anti-aging" is near-useless even if it works as hoped
05:15 < fenn> this at least has a chance of mattering
05:16 < fenn> i don't understand why rapamycin is the baseline on which they're adding more interventions
05:17 < hprmbridge> Eli> Good question.
05:17 < fenn> i get that there's a combinatorial explosion but why not like, control, all the things we think should work, plus new intervention, all new interventions, all new and old interventions
05:18 < fenn> something like that
05:18 < fenn> and if it works with everything all at once then you have can whittle it down later
05:19 < fenn> s/have//
05:20 < fenn> aubrey was a computer science grad originally, he should know this
05:20 < hprmbridge> Eli> It seems like people have converged on mtor being a major component to health/life extension. The theory is that you can pulse rapa without the typical side effects.
05:24 -!- TMM [hp@amanda.tmm.cx] has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
05:24 -!- TMM [hp@amanda.tmm.cx] has joined #hplusroadmap
05:56 < hprmbridge> kanzure> why isn't caterpillar construction equipment more modular instead of monolithic unibody(ish) machines
05:56 < hprmbridge> kanzure> it's probably all the same hydraulics architecture between machines
05:57 < L29Ah> you can plug in lots of different tools to a tractor, including hydraulics
06:03 < L29Ah> ok maybe not to a caterpillar
06:04 < L29Ah> our construction company has a Case 580G and lots of stuff was plugged at some times
06:04 < hprmbridge> kanzure> i mean wheel base and other things. more like a reconfigurable ant body.
06:10 < fenn> welded joints or continuous cast sections are much stronger and more rigid than bolted connections
06:11 < hprmbridge> kanzure> could you do hydraulic connection
06:11 < fenn> sure, but it will be more expensive and heavier
06:11 < hprmbridge> kanzure> well i wouldn't want my heavy construction equipment to be too heavy....
06:12 < fenn> you wouldn't
06:27 < hprmbridge> kanzure> oh right, the david gerard article that i use to show some FTX / EA connections is the same david gerard that the rat community dislikes due to various long-term wikipedia camping grudge moderator issues https://news.ycombinator.com/item?id=44882138
06:30 -!- justanotheruser [~justanoth@gateway/tor-sasl/justanotheruser] has joined #hplusroadmap
06:35 < hprmbridge> kanzure> other than wikipedia edit war issues and the castle built of sand that wikipedia is, not sure what i am supposed to get out of this drama; am I meant, by the HN reply, to discard all his opinions because of a history of involvement in wikipedia trench warfare? and/or because of possible history of Being Wrong? so therefore his FTX/EA article is also what, completely wrong?
06:35 < hprmbridge> kanzure> am I supposed to be surprised that the author of an article has a history of engaging online in those topics?
06:38 < hprmbridge> kanzure> "unreliable, editorially incompetent, repeatedly caught publishing false information, conspiracy theories and hoaxes, [undue weight] for opinions.” ok
06:45 < hprmbridge> kanzure> https://biostasis.substack.com/p/the-false-claim-of-cryonics-as-pseudoscience
07:13 < hprmbridge> kanzure> "it seems like this is a cargo-cultish transposition of the act of finding _fallacies in arguments_ into the domain of finding _faults in persons_." well, people can be faulty and unreliable, but yes the two should not be confused.
07:21 < hprmbridge> kanzure> "...  "steel manning" betrays a lack of charity [or purpose]. If the only thing you owe your interlocutor is to use your "prodigious intellect" to restate their own argument in the way that sounds the most convincing to you, maybe you are in fact a terrible [participant who is not actually committing to update your beliefs, actions or plans and instead are just wasting everyone's time]." actually
07:21 < hprmbridge> kanzure> i used to see steelmanning mainly used as a way to de-escalate violent arguments and offer a branch for good faith engagement, but i suppose this requires both participants to understand the human purpose or moral value of argument beyond entertainment or stress relief.
07:23 < hprmbridge> kanzure> probably the confusion of faults-in-arguments into faults-in-persons is because there is no consensus acceptable way to otherwise reveal or act upon associative preferences?
07:26 -!- stipa_ [~stipa@user/stipa] has joined #hplusroadmap
07:26 < L29Ah> https://americanaffairsjournal.org/2021/08/the-value-of-nothing-capital-versus-growth/ on stocks and memecoins
07:28 -!- stipa [~stipa@user/stipa] has quit [Ping timeout: 276 seconds]
07:28 -!- stipa_ is now known as stipa
07:28 < kanzure> how is there still no "wikipedia 2" plan, or at least a "here is everything that went wrong with wikipedia please do not repeat our mistakes" document
07:30 < jrayhawk> the same reason bitcoin doesn't have a code of conduct
07:32 < kanzure> yeah i keep telling bitcoin core maintainers to delete the code of conduct https://github.com/bitcoin-core/meta/blob/main/MODERATION-GUIDELINES.md
07:34 < kanzure> in the era of LLMs possibly we need better models of factual information extraction and documentation other than a crude digitization of the concept of an encyclopedia
07:34 < L29Ah> semantic web strikes again
07:35 < kanzure> that's more in the platonic-ideal realm tho; even just a monarchist BDFL approach to a wikipedia v2 could have significantly better outcomes even if not meeting any sort of platonic ideal.
07:36 < kanzure> which then becomes a question of bootstrapping adoption, momentum, and delivering greater value than wikipedia
07:37 < L29Ah> there are several wikipedia forks that you never hear about
07:38 < kanzure> L29Ah: just as kimi k2 and your arxiv link yesterday use extensive LLM pre-processing of training data to break down training data into additional steps (etc), the same should be done from an epistemological perspective, because tracing knowledge or the origins of knowledge is probably of greater value to training a good LLM than memorizing matters of fact or history.
07:38 < kanzure> L29Ah: are any of those wikipedia forks any good?
07:38 < L29Ah> idk
07:39 < L29Ah> off the top of my head i can only name http://cyclowiki.org that has a few articles that aren't suitable for wikipedia proper yet still have some value
07:39 < L29Ah> no idea re governance
07:42 < L29Ah> https://wikipediocracy.com wikipedia critics
07:42 < L29Ah> https://wikispooks.com/wiki/Wikipedia/Problems wiki wikipedia criticism
07:43 < L29Ah> http://www.wikiznanie.ru another russian wikipedia fork
07:44 < L29Ah> i don't think i ever seen anything else notable on topic
07:53 < fenn> tracing origins is orders of magnitude more work
07:54 < kanzure> well, i mean origins even within the prompt window
07:54 < kanzure> er, context window
07:55 < fenn> yep
08:29 < kanzure> https://github.com/EIT-NLP/Awesome-Latent-CoT latent chain-of-thought reasoning papers.
09:44 < kanzure> we should explicitly conduct a search for high intelligence (high IQ) humans that specifically score low on IQ polygenic trait scoring.   or i suppose also interesting would be scoring high and being actually low IQ (confirmed by family study, not just bumped the head a little too hard that one time).
09:47 < kanzure> this won't work because UK biobank and other large genome repositories are currently very "skewed" data sources, so you can cheat by going to a different ethnic group that hasn't been as extensively sampled or included in the polygenic trait scoring algorithm training data :(
11:08 <+gnusha_> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=b8c368bb Bryan Bishop: SPOCK1 mathematics ability and more DUFF1220 olduvai domain stuff >> http://diyhpl.us/diyhpluswiki/nutrition/
11:17 < kanzure> if you can cause ancephalic human fetal development via embryo engineering, then i wonder if you could introduce an engineered non-ancephalic neural cell line into the embryo and have it develop the brain instead (modulo some oversimplification here as to how the tissues develop and differentiate into the rest of the anatomy that i don't entirely remember at this moment). this way, you ...
11:17 < kanzure> ...wouldn't have to deal (as much) with non-brain pleiotropy.
11:19 < kanzure> or the opposite: myocytes or muscular tissue shouldn't have to be running on the same genomic baseline program as neurobiology in the same organism.
11:40 <+gnusha_> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=35f3715e Bryan Bishop: put Cas9 into the genome to make it available for expression for future gene therapy >> http://diyhpl.us/diyhpluswiki/genetic-modifications/
11:53 <+gnusha_> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=d4f403e3 Bryan Bishop: other various germline updates, maintenance >> http://diyhpl.us/diyhpluswiki/genetic-modifications/
12:01 < kanzure> "relational training" (something about difference comparisons for more/less or bigger/smaller?) apparently confers +5 to +9 non-verbal IQ points.
12:35 < kanzure> "artificial parthenogenesis" https://embryo.asu.edu/pages/artificial-parthenogenesis-and-fertilization-1913-jacques-loeb "... Silkworm experimenters in the mid-1800s had shown that exposing silk worm eggs to sulfuric acid or rubbing them lightly with a brush could induce parthenogenesis."
12:37 < kanzure> eugène bataillon reported "normal parthenogenetic tadpoles in Rana temporaria" induced by "parthénogenèse traumatique"
15:16 -!- darsie [~darsie@84-113-82-174.cable.dynamic.surfer.at] has quit [Ping timeout: 255 seconds]
16:03 < pasky> kanzure: from the bird's eye point of view, wikipedia is / has been a great success and *not that many things* went wrong, tbh, even though you could put together an endless list of minor things that did
16:03 < hprmbridge> kanzure> so?
16:04 < pasky> so it's not clear what should go on that list, in the sense of prioritization .. because there isn't really much "potential unfulfilled, and root cause is X"
16:05 < pasky> (I also do think wikipedia is likely going to kinda crash and burn in the next 5 years, but that's yet to happen)
16:05 < L29Ah> why would it?
16:06 < L29Ah> i can only name technological inefficiency but this part is handled well by the generous sponsors apparently
16:07 < pasky> llms will wreak havoc on it and/or make it obsolete and i suspect the community is too ossified to react effectively
16:08 < L29Ah> do you mean automated conversion into ads and political propaganda?
16:48 < hprmbridge> kanzure> https://gpupuzzles.answer.ai/intro
16:49 < hprmbridge> kanzure> https://leetgpu.com/
17:22 < L29Ah> no vulkan >:O
18:26 < hprmbridge> kanzure> turbomachinery tool https://github.com/OpenOrion/turbodesigner
20:10 -!- pasky [~pasky@nikam.ms.mff.cuni.cz] has quit [Server closed connection]
20:10 -!- pasky [~pasky@nikam.ms.mff.cuni.cz] has joined #hplusroadmap
20:24 -!- TMM [hp@amanda.tmm.cx] has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
20:24 -!- TMM [hp@amanda.tmm.cx] has joined #hplusroadmap
22:04 < hprmbridge> kanzure> "no for real $8 on primeintellect can get you over 100 million tokens on qwen3 480b coder at 768k context" chat is this true? he is claiming 25,000 tokens/sec on h200
23:34 -!- darsie [~darsie@84-113-82-174.cable.dynamic.surfer.at] has joined #hplusroadmap
23:57 -!- WizJin_ [~Wizzy@2402:a00:184:8354:55f:6a75:8aa4:5e93] has quit [Quit: Leaving]
23:57 -!- WizJin [~Wizzy@2402:a00:184:8354:55f:6a75:8aa4:5e93] has joined #hplusroadmap
23:57 -!- WizJin [~Wizzy@user/WizJin] has changed host
--- Log closed Thu Aug 14 00:00:10 2025