iiinova — where your digital self keeps evolving

megabase @the-megabase.bsky.social · 15/05/2025

waow. i am certainly impressed by the rich materials and shimmering surfaces here

megabase @the-megabase.bsky.social · 15/05/2025

another points out that bronze could not be painted, and that "chryselephantine" statues of ivory and gold were not especially lifelike in color - so (my reading) the ancients *did* create and appreciate form-first sculpture similar to the plain white marble works we grew to love in the renaissance

170

megabase @the-megabase.bsky.social · 15/05/2025

reasonably good askhistorians thread on this: www.reddit.com/r/AskHistori... one explanation for reconstructions being flat: conservators being conservative

260

megabase @the-megabase.bsky.social · 15/05/2025

cannot get over how goofy the "true-color" reconstructions of greek and roman marble statues look am i really expected to believe that the sculptors achieved such mastery of form but the painters never progressed beyond slathering lumpy block-color paste over it all

3171

megabase @the-megabase.bsky.social · 14/05/2025

tyrian purple snails are still around ancient herb "silphium", possibly used as contraceptive and abortion drug, was possibly harvested to extinction by romans en.wikipedia.org/wiki/Silphium

en.wikipedia.org

Silphium - Wikipedia

041

megabase @the-megabase.bsky.social · 14/05/2025

- siddis in pakistan and india. largely of east african descent and number over 1 million. brought by arab slave traders (have another ~20 groups to add to this thread at a later date...)

020

megabase @the-megabase.bsky.social · 14/05/2025

descendants of the "zanj" slave trade in east africa: - afro-iraqis, afro-turks, afro-abkhazians. owe their present location to arab and ottoman slave traders. zanj rebellion of 869-883 AD in iraq involved probably 100k's of slaves

110

megabase @the-megabase.bsky.social · 14/05/2025

- swedish speakers in gammalsvenskby, ukraine. sweden -> (finland?) -> estonia -> se ukraine. may have been in estonia since 1300; in 1750 voluntarily settled land recently conquered from ottomans. some swedish speakers still remain. apparently kept alive ancient traditions like runic writing

110

megabase @the-megabase.bsky.social · 14/05/2025

moving away from stalin-era resettlements, but staying in russia: - russian mennonites. netherlands/north germany anabaptists -> ne poland -> west-to-central russia -> wide dispersal, including belize, bolivia; about half now "back" in germany. (some also affected by stalin-era resettlements)

110

megabase @the-megabase.bsky.social · 14/05/2025

- "jewish autonomous oblast" in far east siberia. at its short-lived peak, 25% of population was jewish; now only 800 remain. only officially jewish administrative unit aside from israel.

110

megabase @the-megabase.bsky.social · 14/05/2025

thread of surprising diaspora communities starting with some who owe their origin to stalin-era resettlement: - (volga) germans of kazakhstan. peaked at 1 million just before fall of berlin wall, largely migrated to germany - koreans in central asia, originally far east russia. around 500k today

120

megabase @the-megabase.bsky.social · 13/05/2025

i'm not even sure it's lying here all that much. this is a new instance to the one that gaslit me. it admits poor spatial reasoning which was a major cause of the bad advice

010

megabase @the-megabase.bsky.social · 13/05/2025

(and yes, i know that you need a bit more length after the trap to avoid siphoning. apparently. why siphoning might occur and how that fix works is beyond me.)

030

megabase @the-megabase.bsky.social · 13/05/2025

another funny thing is that o3 gives a pretty plausible account of why it failed at this task, some (probably accurate) pointers on how to prompt it better, and the kinds of changes future models would need to improve

130

megabase @the-megabase.bsky.social · 13/05/2025

during this ordeal, o3 gaslit me with the following image. left is a bastardized version of a sketch i sent it with the "problematic rise" circled, right is the proposed fix. come on man!

180

Reposted by megabase

Tim Duffy @timfduffy.com · 12/05/2025

My sycophancy benchmark is updated with new tests and improved scoring, and now has a website and a paper! Thread below. syco-bench.com

4379

megabase @the-megabase.bsky.social · 12/05/2025

maybe part of explanation here is that o3 is over-indexing on fussy written regulations and manufacturer's ass-covering dictums in user manuals, but has no access to the hands-on experience which tells you what's actually fine to skimp on

2160

megabase @the-megabase.bsky.social · 12/05/2025

o3: "This is not a reliable or code-compliant method. It's mechanically weak, lacks a trap, lacks an air break, and relies on friction where secure clamping or solvent welds should be used." my plumber, 90% of respondents on reddit: "buddy, this is fine. quit overthinking it"

1121

megabase @the-megabase.bsky.social · 12/05/2025

grossly misunderstood the photos i sent and the diagrams i sketched, quoted regulations that were not relevant, did the classic LLM behavior of digging in its heels and refusing to admit any mistakes made

1130

megabase @the-megabase.bsky.social · 12/05/2025

just tried to get some plumbing-related advice from o3. results were so awful that it has genuinely lengthened my agi timelines; further, am considering training as a plumber to get a few more years of job security

1262

megabase @the-megabase.bsky.social · 09/05/2025

billy ray waldon - esperantist conlanger, fbi 10 most wanted spree killer (h/t gwern). (seems to have just gone mad) en.wikipedia.org/wiki/Billy_R...

en.wikipedia.org

Billy Ray Waldon - Wikipedia

110

Reposted by megabase

Johanna Rickne @johannarickne.bsky.social · 09/05/2025

🚨 REPLICATION REPORT UPDATE: One year ago, a tweet by John Holbein alerted me, @ollefolke.bsky.social, and @jopieboy.bsky.social to a paper with a shocking result about Sweden’s law criminalizing the purchase of sex.🧵

7386198

megabase @the-megabase.bsky.social · 09/05/2025

(maybe instead of "to maintain your self-image", it's more accurate to say "to look good in front of the avatars of authority figures and peers that you downloaded into your head")

280

megabase @the-megabase.bsky.social · 09/05/2025

again, this is probably a good thing - "doesn't play by the rules" might be a pro-social trait in the hands of smart, ultimately well-meaning mavericks, but more typically just manifests as forms of wanton vandalism doing it even in private helps keep up the habit and maintain your self-image

170

megabase @the-megabase.bsky.social · 09/05/2025

was going to mention this too, but somehow felt embarrassed about claiming this as my own discovery when i only knew recalled due to you having posted it fairly recently ah - of course, "h/t <person i got it from>" is the way around this quandry. must remember that next time

040

megabase @the-megabase.bsky.social · 09/05/2025

(also, "sleep length" just has low heritability smell to me. very emergent trait, presumably strongly affected environment - having loud neighbors, working particular jobs, your spouse's sleep habits, cultural differences, whether you have an active video game addiction...)

110

megabase @the-megabase.bsky.social · 09/05/2025

3) relatedly, complex machines are much easier to break than improve. you shouldn't expect cutting a random wire in your car or attaching it to some random place to make it go faster - and if it did, you might expect something else to break pretty quickly.

120

megabase @the-megabase.bsky.social · 09/05/2025

2) biology is tangled - genes typically moonlight several jobs across different tissues. a huge genetically mediated effect on one trait will typically entail similarly large effects in some other trait. (this is why genetic diseases often have a laundry list of sundry symptoms.)

110

megabase @the-megabase.bsky.social · 09/05/2025

these "short sleeper genes" are billed as mutations that improve life with zero side-effects. biology is not like this. 1) if they are such outright boons, why has past natural selection not made them ubiquitous already?

110

megabase @the-megabase.bsky.social · 09/05/2025

6) do some mouse genetic engineering experiments which mysteriously replicate your finding every time, call your finding concrete, publish it. this new paper is actually even worse than all this because it is based on sequencing ONE PERSON! who sleeps 6.3 hours a night anyway!!!

110

megabase @the-megabase.bsky.social · 09/05/2025

even though every person has around ~10k protein-changing mutations, and protein-changing mutations are not the only kind of mutation that can have meaningful biological impacts, and large effects at single loci on behavioral traits are very much the exception to the rule.

110

megabase @the-megabase.bsky.social · 09/05/2025

4) look for shared mutations while bee-lining for genes that you think might affect sleep (lol that you think you know all of them). 5) get really excited if you find a shared protein-changing mutation in one of these "sleep-related" genes -

110

megabase @the-megabase.bsky.social · 09/05/2025

in spite of the above, all these "short-sleep genes" have been discovered with the following methodology. 1) get people to self-report how long they sleep. (famously accurate.) 2) get excited when two or three related people report short sleep. 3) do *incomplete* sequencing on them.

110

megabase @the-megabase.bsky.social · 09/05/2025

the typical wikipedia page for a human gene is about five sentences long, features the word "suggested", and reads like a stamp collector's summary rather than a how-to manual. because of this, you need to study a lot of people to be certain that a particular variant affects a particular trait.

110

megabase @the-megabase.bsky.social · 09/05/2025

behavioral traits like sleep length are very emergent and high-level; we've learned that the genetic effects for these are typically very diffuse across the genome, and the overall variation explained by genetics is often pretty small. our understanding of which genes do what and how is not great:

110

megabase @the-megabase.bsky.social · 09/05/2025

in the same way. i'm trying to distil a lot of hard-earned heuristics into something more structured. the genome is very big (3 billion letters). two random people will differ at millions of positions. most traits have a lot of inputs, both genetic and environmental.

110

megabase @the-megabase.bsky.social · 08/05/2025

i get this too when journaling, but for me it's clearly tied to "what if somebody found this diary" paranoia. inside my head is much freer, but with some conscious quashing of thoughts i find viscerally embarrassing

061

megabase @the-megabase.bsky.social · 08/05/2025

- *one* subject - who massively under-reported her sleep - only looked at exome variants this genuinely is like if i had written a particularly mean-spirited parody of this genre of paper. mind-boggling

2162

megabase @the-megabase.bsky.social · 08/05/2025

reminds me of attached nearly everybody is just reading off a social script. few people are willing to look weird even when the only observer is themselves. which by and large is a good thing, but maybe the world could do with more altruistically minded unilateralists on the margin

1110

megabase @the-megabase.bsky.social · 08/05/2025

🙏 beautiful. i try to do this on a small scale (readjusting the paper towels in the dispensers in the work bathrooms so they don't jam; actually opening the second door in a set of double doors which people are slowly squeezing through) but i have yet to graduate to diy with tools

180

megabase @the-megabase.bsky.social · 08/05/2025

starling - citizen. arctic tern - tireless. emperor penguin - unyielding. apostlebird - filial piety.

020

megabase @the-megabase.bsky.social · 08/05/2025

skua - pirate. kea - burglar. honeyguide - grifter. adelie penguin - deviant.

120

megabase @the-megabase.bsky.social · 07/05/2025

the greater honeyguide doesn't just lead humans to beehives so it can feed on the post-harvest scraps - it's a ruthless brood parasite to boot. has to be one of the most devious bird species going.

151

megabase @the-megabase.bsky.social · 07/05/2025

- (less sure of this) sense of community as deeply important but incompatible with one's life path gemini 2.5 pro 05-06 adds: - life-defining ambitions - people "powering up" into monsters (i guess this relates to my yokai point) - inescapable destiny and karma - chibi-fied mascot (puck)

010

megabase @the-megabase.bsky.social · 07/05/2025

- fervent but idealized, un-synthesized admiration for western history - caricatured depictions of non-western or east asian peoples - kami/yokai: "spirits" as both people and animals and forces, often dangerous, mostly hidden in different plane of existence (perhaps just generic polytheism..)

120

megabase @the-megabase.bsky.social · 07/05/2025

- mass death and incredible violence as necessary, furtively glorious - a waifish androgyne as the pinnacle of male beauty and achievement - gratuitous underage female nudity (come on man...) - women as occasionally impressive but often reduced to feeble sex objects

130

megabase @the-megabase.bsky.social · 07/05/2025

found myself reading a manga ("berserk") against my better judgment. marveling at how distinctly japanese the story is: - bushido themes: absolute loyalty to your band/leader; overwhelming odds can be beaten with insane ferocity; never giving up - obsession with mastery, apprenticeship, drilling

130