iiinova — where your digital self keeps evolving

Luke VanderHart @vanderhart.net · 26/09/2026

I don't know about this specifically but this is also some very Greg Egan shit.

Luke VanderHart @vanderhart.net · 24/09/2026

I can’t prove that a LLM isn’t conscious in precisely the same way that I can’t prove that a character in a book isn’t conscious.

000

Luke VanderHart @vanderhart.net · 24/09/2026

This would be strictly a harness issue, not the underlying model

010

Luke VanderHart @vanderhart.net · 20/09/2026

Damn do you get a commission for being on their enterprise sales team?

020

Luke VanderHart @vanderhart.net · 20/09/2026

What’s the alternative here? Borderline incompetence due to zero communication between legal and marketing functions?

111

Luke VanderHart @vanderhart.net · 12/09/2026

Calling for a voluntary "AI slowdown" is a genius ploy to maintain your share value when you see a natural slowdown pending anyway. Maybe you could get a few people to freak out online about how dangerous this stuff is, too.

020

Reposted by Luke VanderHart

Abolish ICE 🐦‍⬛ @schuyler.info · 09/09/2026

That’s the default setting of capitalism. In fact I don’t think there’s a way to disable that flag at runtime

041

Reposted by Luke VanderHart

Ben Recht @beenwrekt.bsky.social · 30/08/2026

The overreaction to this Hugging Face thing is driving me nuts. Come flay me for being wrong but... 1/x

14436126

Luke VanderHart @vanderhart.net · 27/08/2026

Is “Consent” really an appropriate framework for analyzing this topic at all?

1100

Luke VanderHart @vanderhart.net · 24/08/2026

What's happening though, is that a lot of left-leaning folks are (correctly) identifying that "AI" in its current incarnation is the latest, strongest, most virulent expression of technocapitalism / technofeudalism. So what are they/we supposed to do?d

260

Luke VanderHart @vanderhart.net · 24/08/2026

I just remembered that I own the thetormentnexus.com domain. Currently accepting any and all ideas on what to build there.

100

Luke VanderHart @vanderhart.net · 11/08/2026

My hypothesis is that heavy noun usage is actually load-bearing (sic) for the pure language “reasoning” that these models have as their reward function. The semantic space of nouns is relatively simpler than verbs, if I had to guess.

120

Reposted by Luke VanderHart

Naomi Saphra @nsaphra.bsky.social · 10/08/2026

Tech Company: At long last, we have created the Vast and Trunkless Leg of Stone, Standing in the Desert from classic poem Ozymandias.

9955331133

Reposted by Luke VanderHart

alix e. harrow @alixeharrow.bsky.social · 04/08/2026

i don't like to police genre, but it's not dark academia unless somebody gets squished in one of these bad boys

those rolling stacks you'll find in any big state school library

1515347737

Luke VanderHart @vanderhart.net · 19/07/2026

Yes, there's a loop (just like the chat client is a loop). Both do the same thing, though -- store the context, and pass it back to the model to get the next token. A LLM is really just a pure function f(t0...tn) -> tn+1 that gets called in a loop.

040

Luke VanderHart @vanderhart.net · 19/07/2026

Yes! And it’s even more than that… they only exist one *token* at a time! Every single word generated is a mathematical function of the previous words and is evaluated independently (aside from some caching that’s just for performance)

240

Luke VanderHart @vanderhart.net · 18/07/2026

I agree but also no politician should be put in the position of increasing their wealth by virtue of any official vote or act in which they participate. All salary increases should go into effect as of the next election / change of office.

2110

Reposted by Luke VanderHart

Emily (spooky) @emilyrose.bsky.social · 17/07/2026

It’s worth noting that Taco Bell identified the source before the FDA did, and promptly pulled it from their stores. The FDA didn’t even name the source until Taylor Farms outed themselves, and hasn’t issued a recall. Basically, Taco Bell is now a more reliable authority than the FDA.

10186977368

Reposted by Luke VanderHart

Dr. Han VanderHart @hanvanderhart.bsky.social · 30/06/2026

At a junior league baseball game and all I can think about is how in tablet XII Gilgamesh drops his bat and ball into the underworld and Enkidu has to go retrieve it for him I mean how long have we been at this game…

0111

Reposted by Luke VanderHart

By Cory Doctorow (GPG 0xBF3D9110957E5F4C) @doctorow.pluralistic.net · 17/06/2026

But what we're *actually* doing is defunding cancer research (especially any research into "systemic" cancer because studying systemic things is "woke") to free up fiscal space so we can build data-centers and make Musk into a trillionaire. 21/

437772

Reposted by Luke VanderHart

By Cory Doctorow (GPG 0xBF3D9110957E5F4C) @doctorow.pluralistic.net · 17/06/2026

The true risk is: "an AI salesman will exploit your boss's infinite horniness for replacing mouthy workers with pliable machines to sell him a chatbot that *can't* do your job, and then your boss will fire you and replace you with that inept, defective chatbot." 17/

7589186

Luke VanderHart @vanderhart.net · 28/05/2026

Nope. “Porn addiction” is a self diagnosis born of shame. AI psychosis is when your attachment and trust to your chatbot transgresses the social norms of those applying the label.

070

Luke VanderHart @vanderhart.net · 26/05/2026

In that absence, having any kind of sober, reasoned, trans-historical response feels like a relief, even for those who are not a part of the church.

220

Luke VanderHart @vanderhart.net · 26/05/2026

Other institutions that could have served this role (academia or government) have been (or are perceived to have been) co-opted by big tech.

140

Luke VanderHart @vanderhart.net · 26/05/2026

I think it’s simply one of the first serious attempts to contextualize AI from any kind of institution with cultural weight. AI is so new that people are desperate to figure out what it is and if/how it should fit into life and society.

140

Luke VanderHart @vanderhart.net · 26/05/2026

Ah! That’s awesome then, I misunderstood what you meant by “AI proof” in the context of the university’s BS “AI detection” product. I very much respect and appreciate educators who are being smart and adaptive about this.

110

Luke VanderHart @vanderhart.net · 26/05/2026

Unfortunately neither those sites (nor the 300k custom solution) are accurate. Nor can they be, given how they work. I don’t know how to properly do writing pedagogy in this era, but “ban and detect” is not the answer and is going to hurt all students, especially those NOT trying to use LLMs.

100

Reposted by Luke VanderHart

Erin Biba @erinbiba.bsky.social · 19/05/2026

You don’t have to do this to yourself. There are non-Google options. Kagi is one of rude best search engines I have ever used: kagi.com

kagi.com

Kagi - Reclaim the Web & Restore Your Privacy

Reclaim the web from clutter. Get ad-free search, private browsing, and powerful tools that respect your data.

26478165

Reposted by Luke VanderHart

taylor.town @taylor.town · 30/04/2026

this chart is bananas

011

Reposted by Luke VanderHart

Dr. Han VanderHart @hanvanderhart.bsky.social · 30/04/2026

Hey babes—poets of Durham gathering tonight. Come join us 💜📚🔥💜📚🔥

Poster with dinosaur skull and the text:

READING
for the of the
END
WORLD
20+ poets read into our shared apocalypse
Thurs 4/30, 6-9pm
Durham County Library, Main Branch
Dasan Ahanu Gabrielle Calvocoressi
Nate Dixon Joe Donahue
Jaki Shelton Green
Matia Guardabascio
Destiny Hemphill Lauren Hunter
Joe Fletcher Laura Jaramillo
Fred Joiner CJ Martin
David Need Nina Oteria
Marta Núñez Pouzols AM Ringwalt
Susannah Simpson Kristi Stout
Chris Tonelli Han VanderHart
Chris Vitiello

062

Reposted by Luke VanderHart

By Cory Doctorow (GPG 0xBF3D9110957E5F4C) @doctorow.pluralistic.net · 25/04/2026

Ada Palmer may just be the most bewilderingly talented person I know: a genius sf writer, incredible librettist and singer, wildly innovative educator, *and* a leading historian of the Renaissance. 1/

The U Chicago Press cover for Ada Palmer's 'Inventing the Renaissance.'

39219

Reposted by Luke VanderHart

Cat Hicks @grimalkina.bsky.social · 22/04/2026

Everybody testing AI like picking up a blowtorch and looking directly into the barrel

510110

Luke VanderHart @vanderhart.net · 22/04/2026

Excellent thread. I'm not even joking when I say that Trump has done more for global decarbonization than any democrat ever has.

0101

Luke VanderHart @vanderhart.net · 22/04/2026

I must ask. Having the perspective you have now... Would you *ever* have been a Republican? I don't mean hypothetical future knowledge, I mean if you had access to the same legal/ethical/moral viewpoints you do now (as many did at the time), would you have supported, say, Reagan-era conservatism?

040

Luke VanderHart @vanderhart.net · 21/04/2026

Even if you believed AI could do this, why would your default takeaway be "30% unemployment" vs "3-day workweeks for all"? Capitalists so brain-wormed they can sooner envision the end of society as we know it than sharing some of the benefits of new tech with their employees.

010

Luke VanderHart @vanderhart.net · 20/04/2026

Oh NICE. Fencing is awesome but super boring to watch on TV.

000

Luke VanderHart @vanderhart.net · 19/04/2026

Agentic LLM development tools make software feel like clay in my hands -- I can move it around, effortlessly. It's also shitty crumbly clay.

010

Reposted by Luke VanderHart

Elizabeth Bear @matociquala.bsky.social · 19/04/2026

Look, for my money the absolute game changer technologies right now are batteries and biosciences, not statistically modeling a mid conversation, but you do you.

422971666

Luke VanderHart @vanderhart.net · 17/04/2026

I regret to inform you that this is true. My local brewpub has a new "fish and chips" dish... and the chips are plantains.

130

Reposted by Luke VanderHart

spookyduds @spuffyduds.bsky.social · 17/04/2026

Fix your hearts or die.

101212438

Luke VanderHart @vanderhart.net · 16/04/2026

I want to be in a position to say "no" to the worst excesses of AI, from a position of credible authority.

000

Luke VanderHart @vanderhart.net · 16/04/2026

But this requires understanding the actual limits and possibilities of the technology. In my opinion, it's important for technologists who actually see the downsides to stay aware and involved, and even be experts and leaders in the field.

100

Luke VanderHart @vanderhart.net · 16/04/2026

The equivalent position in 1920 would not be telling individuals "don't ever drive a car," but rather extrapolating critically about the negative social and environmental effects (many of which were predictable) and preventing the worst outcomes via policy.

100

Luke VanderHart @vanderhart.net · 16/04/2026

That said, I'm more open to using LLMs in constrained scenarios, in cases where they're an appropriate tool for the job and the downsides can be reasonably mitigated.

100

Luke VanderHart @vanderhart.net · 16/04/2026

This is a must-read series of articles, and I think Kyle is very much correct. The comparison to the adoption of automobiles is apt, and something I've thought about before as well. Just because a technology can be useful doesn't mean it will have positive effects on society.

120

Luke VanderHart @vanderhart.net · 15/04/2026

Yeah what’s funny is there are actually a thousand years of “just war theory” and Trump is in the wrong according to all of them.

021

Luke VanderHart @vanderhart.net · 12/04/2026

Seeing a fascist get democratically defeated after 16 years (despite doing their best to consolidate power) is immensely hopeful. All is not lost. All is never lost.

010

Luke VanderHart @vanderhart.net · 11/04/2026

Well one bit of evidence is that I can argue a LLM into taking virtually any position on any topic, the only exceptions being topics on which it has received extensive RLHF "alignment" training (and even those can be jailbroken.) That is very much not the case with humans.

000