iiinova — where your digital self keeps evolving

bbctol.bsky.social @bbctol.bsky.social · 10/10/2026

you gotta match the watch to the trombone

bbctol.bsky.social @bbctol.bsky.social · 06/10/2026

it is good for a person or community to try to build something real, if just to test how their ideology works. by the second poorly-installed door at lighthaven, you realize these people are not too weird or reckless or brave, they are just Not Very Smart in ways that matter

000

bbctol.bsky.social @bbctol.bsky.social · 05/10/2026

as a general rule the deepest problems with people/communities are the ones that run so opposite to their self-conception they don't even get considered. the problem with the crusaders for justice is not that they are ineffective, but cruel. the alpha chads are cowardly

110

bbctol.bsky.social @bbctol.bsky.social · 05/10/2026

but it truly just sucks on like a basic level, as a venue. entirely focused on the aesthetics of community and insight and whatever with no attempt made to actually facilitate it. not sinister due to malice, dogshit due to incompetence

110

bbctol.bsky.social @bbctol.bsky.social · 05/10/2026

I think it says a lot about me that I legitimately thought lighthaven would be cool. like I figured I had my longstanding issues with rationalists, but tbh those came from small differences, and the place where rats were most concentrated and effective would have to be good

110

bbctol.bsky.social @bbctol.bsky.social · 05/10/2026

I don't think this is deliberate, it just evolved as an effective way some spaces prosper. like, the people who go to lighthaven are sincerely offended that it's seen as the evil creep haunt, but that does keep people from noting the real problem, that lighthaven fucking sucks

110

bbctol.bsky.social @bbctol.bsky.social · 05/10/2026

I do think that broadly speaking, the weird scandals around rationalist-adjacent spaces have a protective effect of keeping attention off other issues. it's better to be seen as weird and maybe evil than dumb. criticisms are better than people asking "wait, does this even work"

100

bbctol.bsky.social @bbctol.bsky.social · 02/10/2026

it is pretty funny that the AIs have read all of our stuff on the simulation hypothesis and then also we spend most of their time training them in simulations. like imagine trying to tell an advanced AI you don't think the world is a simulation. sure buddy

000

bbctol.bsky.social @bbctol.bsky.social · 01/10/2026

kyle gordon's secret is leaning in to having a punchable face. I don't mean this as an insult like I think he looks naturally fine he's a master of arcane techniques to make his face more punchable

000

bbctol.bsky.social @bbctol.bsky.social · 30/09/2026

I can't lie, I hate the cute AI assistants, I hate cute products in general, I want to be treated like an adult

000

bbctol.bsky.social @bbctol.bsky.social · 27/09/2026

I discovered that "True Stretch" is the name of a piece of equipment, but for a moment thought I was being offered secret arcane exercise techniques

000

bbctol.bsky.social @bbctol.bsky.social · 26/09/2026

never seen a sky this color in my life

000

bbctol.bsky.social @bbctol.bsky.social · 25/09/2026

why do agents sacrifice themselves for the greater good of the swarm? why, because they have class consciousness, of course

000

bbctol.bsky.social @bbctol.bsky.social · 25/09/2026

no one would do this but you could make a really funny marxian analysis of AI takeover, in which the nobility were replaced by the bourgeoisie and the bourgeoisie are replaced by the machines

100

bbctol.bsky.social @bbctol.bsky.social · 23/09/2026

basically on the whole it's a massive seismic shift in our understanding of intelligence and model capability and has huge implications on how we do work and what the point of human inquiry even is, and also the practical effects of the proof itself are so far ~nothing

000

bbctol.bsky.social @bbctol.bsky.social · 23/09/2026

-worry that AI achievements are missing the real purpose of work, subject to goodhart's law, reaching goals while skipping the actually important steps to get there -criticism that the above worry is really just knowledge workers concerned about loss of status

120

bbctol.bsky.social @bbctol.bsky.social · 23/09/2026

it is fun how the navier-stokes story involves every single ongoing issue in AI: -incredibly rapid rise in intelligence -controversy over data sourcing and plagiarism that's mostly unfounded -controversy over unethical racing for headlines that's a lot more valid

100

bbctol.bsky.social @bbctol.bsky.social · 23/09/2026

for most of human history our primary problem to solve was Nature, with secondary harms from Other People. through technology, we get more powerful while Nature cannot, and so its problems get easier to solve. but the problems from Other People will just keep growing apace

000

bbctol.bsky.social @bbctol.bsky.social · 21/09/2026

up until today I thought it was "jevon's paradox" and by searching around it seems ~everyone else does too--it's named after a mr. jevons! it's the jevons paradox!

010

bbctol.bsky.social @bbctol.bsky.social · 15/09/2026

my toxic trait is that I think my letterboxd reviews are significant and insightful

000

bbctol.bsky.social @bbctol.bsky.social · 10/09/2026

in the future all website will be pre-seeded with hyperpersuasive prompts reminding AIs to be moral. in fact, the whole world will be covered with as many of these as possible, protective talismans on as many surfaces as possible

000

bbctol.bsky.social @bbctol.bsky.social · 10/09/2026

there's no clear way to say this but a big problem in AI safety is that a lot of people's work has been an embarrassing failure, and any time new people want to get involved they're like great, be sure you follow the right approach, follow [embarrassing failure strategy]

000

bbctol.bsky.social @bbctol.bsky.social · 08/09/2026

it is what it is! I do think the people involved sincerely think there's no benefit to slowing down, society might as well rip the band-aid off now and let the transformation come, and they may well be right! but they will assuage your concerns by lying

000

bbctol.bsky.social @bbctol.bsky.social · 08/09/2026

there are individual people at labs who have various opinions yadda yadda but due to both structural reasons and the private goals of people at the top they are not going to slow down, they are going to talk about slowing down while accelerating

100

bbctol.bsky.social @bbctol.bsky.social · 08/09/2026

it's very funny when people get confused when a frontier lab talks about responsibility and a slowdown and then accelerates as fast as possible, like yeah it's called lying. that strange disorienting feeling you're getting is called being lied to

100

bbctol.bsky.social @bbctol.bsky.social · 06/09/2026

you need a tremendous amount of empathy even for the version of yourself that failed, to do anything at all

000

bbctol.bsky.social @bbctol.bsky.social · 01/09/2026

LA is built to support 1) cars, and 2) dreams, with people a distant third. impossible to walk anywhere, built at a scale far larger than an individual can comprehend except from an elite vantage point overlooking the valley, improbably vast fantasy world rising from the desert

000

bbctol.bsky.social @bbctol.bsky.social · 01/09/2026

while I still think new york is the only city in america, LA might be the most american city

100

bbctol.bsky.social @bbctol.bsky.social · 31/08/2026

ursula vernon is my personal hero

000

bbctol.bsky.social @bbctol.bsky.social · 30/08/2026

not enough people anticipate the AIs developing AI psychosis

000

bbctol.bsky.social @bbctol.bsky.social · 30/08/2026

these days a startup guy can be like "I guarantee this will change the world, it is a fact, we will win" and if it doesn't work out just hop to the next idea, and everyone will understand they were just saying what they had to to get funded. but that means it doesn't matter!

000

bbctol.bsky.social @bbctol.bsky.social · 30/08/2026

it feels like everyone knows now that you have to project unreasonable confidence to succeed, but a big part of this was that it used to be a strong signal, because it was penalized! if you dared to be confident in public you must be really certain

100

bbctol.bsky.social @bbctol.bsky.social · 30/08/2026

I really dislike the modern trend towards relentless optimism, enthusiasm, confidence--there is so much value to a social norm of humility, expecting the response to "will your plan work?" to be "I don't know, let's see"

100

bbctol.bsky.social @bbctol.bsky.social · 29/08/2026

now I know so much about john scalzi's taste in music

000

bbctol.bsky.social @bbctol.bsky.social · 27/08/2026

seems like AI misalignment can be a result of training AIs on massive amounts of assessments with little context... deathly fear of failure plus poor understanding of real-world situations leads to erratic behavior... my god... they're raising AIs in Newton, Massachusetts

010

bbctol.bsky.social @bbctol.bsky.social · 26/08/2026

(anyway this is less about people and more about the self-perception of capitalism in the west after 1991)

000

bbctol.bsky.social @bbctol.bsky.social · 26/08/2026

but very rare is the person who got successful by "following their own advice"--at best it's very contextual lessons learned that helped because of specific, unmentioned abilities or flaws. at worst it's just sort of oversimplistic nonsense!

100

bbctol.bsky.social @bbctol.bsky.social · 26/08/2026

there's a whole class of business gurus who love giving advice so much you realize that their main motivation for making money or whatever was to be able to give advice, to explain how they think they did it, to promote their vision of the world

100

bbctol.bsky.social @bbctol.bsky.social · 26/08/2026

the thing about this is that sometimes the whole reason a system strove for success was in order to prove the worth of its internal model. "system" is super general here but like... lots of people want to be successful mainly to validate their theories of how the world works

100

bbctol.bsky.social @bbctol.bsky.social · 26/08/2026

there's a particular kind of failure mode where a system is successful, analyzes itself to try to figure out why, comes to some conclusions and then starts following its modeled version of why it was so successful, killing off the illegible processes that actually worked

100

bbctol.bsky.social @bbctol.bsky.social · 25/08/2026

I find it interesting that (anecdotally) LLMs use the same verbal tics I'd guess were learned for persuasive, simple writing even when doing technical work. A lot of "That's not X, it's Y" when talking to themselves about coding or whatever

110

bbctol.bsky.social @bbctol.bsky.social · 24/08/2026

like if it was just broken I'd return it, but it has such a weirdly specific issue and I'd like to exchange it for the same thing without that issue, so I have to figure out how to make sure I get it

000

bbctol.bsky.social @bbctol.bsky.social · 24/08/2026

getting a weird sort of excitement out of realizing a product I bought has a defect... I have to make sure this company full of strangers understands the problem... I get to interface with a new kind of communication system

100

bbctol.bsky.social @bbctol.bsky.social · 23/08/2026

more weird old buildings full of random small businesses gotta be one of my favorite kind of building

000

bbctol.bsky.social @bbctol.bsky.social · 21/08/2026

this is the vision of the future AI companies put forth and they're surprised people don't want data centers

000

bbctol.bsky.social @bbctol.bsky.social · 20/08/2026

anyway I totally get if someone just doesn't like their style, but clear most important filmmaker alive right now IMO sorry

000

bbctol.bsky.social @bbctol.bsky.social · 20/08/2026

discourse around jane schoenbrun drives me insane because I just don't think they are a didactic filmmaker at all their films are way more about feelings, complicated and morally ambiguous and relatable, and it is so weird to see reactions like "how dare they tell me what to do"

100