iiinova — where your digital self keeps evolving

I am resigned to NC letting me down, but still…my wife’s hometown, whose whole brand is being pig ignorant and bright red, is “sick of Trump” according to her cousin, who still lives there.

000

Also, the knowledge graph technology actually works, even if it’s never going to remake the world. You can use it to build things humans will use. My thing with “Web3” was that literally no sensible person would have a use for it. Other than maybe criminals.

000

That was the other one, which also didn’t get off the ground, but was less (?) culty. There are a number of knowledge graph folks who are *very* excited about AI, I guess because it means unreliable humans won’t have to be doing the semantic stuff.

100

Does anybody remember them? They were everywhere. I assume they’ve seamlessly grafted themselves into the new AI hotness now.

250

Ethan Gruber @ewg118.bsky.social · 30/09/2026

It’s very much the vibe I got off the Web3/blockchain lunatics a handful of years back. They’re not fully speaking English anymore and they’re convinced they’re now 100x more productive. Could be true, but I don’t join cults, so…

170

Reposted by Hugh Cayless

Depending on how quickly the Library of Congress implements yesterday's EO, both Super Intelligence and Lake America will propagate into the University of Virginia library catalog in about 10 days, unless we take precautions to create a bunch of local headings.

132

Reposted by Hugh Cayless

McSweeney's @mcsweeneys.net · 29/09/2026

"Can a rapist’s prose elicit remorse for sexual assault? What word count best equates to appropriate jail time? In this class, we will dissect what empathy and accountability might look like so undergrads may successfully mimic them on paper in pursuit of avoiding just punishment." buff.ly/5na4oMY

buff.ly

Announcing Cornell University’s Newest Writing Course: Rape Essays 101

“Can a rapist’s prose effectively elicit remorse for sexual assault? What word count best equates to appropriate jail time? In this class, we will dissect wh...

19830273

Hugh Cayless @hcayless.bsky.social · 29/09/2026

Probably going to be an interesting AI principles committee meeting tomorrow, given the ways it’s jumped the shark.

021

Reposted by Hugh Cayless

kingsdigitallab.bsky.social @kingsdigitallab.bsky.social · 17/09/2026

Do you enjoy collaborative, applied work at the intersection of research and technology? We’re recruiting a Senior Research Software Engineer. We work closely with researchers, developing thoughtful, impactful digital research projects, mainly in arts & humanities. www.kcl.ac.uk/jobs/158445-...

kcl.ac.uk

Senior Research Software Engineer | King's College London

168

Hugh Cayless @hcayless.bsky.social · 29/09/2026

Wow. Some bright-eyed aide is getting fired today.

110

Reposted by Hugh Cayless

ve, the ghost of info past @vortexegg.com · 25/09/2026

Incredible scenes from Meta. Apparently you can prompt inject their Muse agent into giving you a root shell on the box it's running on by roleplaying that you are actually the Muse agent itself.

neuromatch.social

jonny (nonvenomous) (@jonny@neuromatch.social)

Attached: 1 image Not only confirming this, its extremely easy. Almost no prompt injection resistance lmao

16375108

Hugh Cayless @hcayless.bsky.social · 25/09/2026

That was my tag line on Twitter (as it was then). I’m relieved that my current job is not forcing us to use AI.

040

Hugh Cayless @hcayless.bsky.social · 24/09/2026

I’m not saying it can’t happen that a noncitizen ends up registered to vote, but there needs to be both stupidity and a bunch of checks failing for it to happen. Basically zero immigrants want to come to the attention of the US government. Especially now.

020

Hugh Cayless @hcayless.bsky.social · 24/09/2026

To add insult to injury, the AI botnets are also polluting our web stats, making it look like we have lots of traffic from places we don’t. Lots of love to Brazil, but there’s not a lot of humans looking at papyri from there.

080

Reposted by Hugh Cayless

jamelle @jamellebouie.net · 23/09/2026

i think you can fairly say that the second trump administration is the most racist federal administration since woodrow wilson and honestly i'm not sure that it isn't more racist than wilson's

164113092110

Hugh Cayless @hcayless.bsky.social · 23/09/2026

This was my first one too. Dad picked it up during a layover in Miami. 16K of RAM! And those wonky joysticks.

030

Reposted by Hugh Cayless

Schrödinger's Sneetch Belly @rtodkelly.bsky.social · 22/09/2026

Please know that this is one of my few legitimate areas of actual, high-level professional expertise: Kicking the healthiest people off of a health insurance program will bankrupt it. Not 'could;' not 'might,' not 'probably," but absolutely will, with mathematical certainty.

25768212249

Reposted by Hugh Cayless

Courtney Milan @courtneymilan.com · 22/09/2026

THIS IS THE KIND OF SHIT I WANT MY GOVERNMENT TO DO Hold large corporations accountable. Distribute the proceeds to the workers. Rinse, repeat.

6879198

Reposted by Hugh Cayless

Hugh Cayless @hcayless.bsky.social · 21/09/2026

Them: “Don’t delete this email!” Me:

041

Hugh Cayless @hcayless.bsky.social · 21/09/2026

Them: “Don’t delete this email!” Me:

041

Hugh Cayless @hcayless.bsky.social · 20/09/2026

I’m agnostic as to whether that’s possible (not yet I’d say), but it depends on how you’re going to define intelligence, and absent an agreed upon definition, we’re just talking past each other. Some people clearly think it’s intelligent, to which I’d ask whether it’s a slave then. Can of worms.

100

Hugh Cayless @hcayless.bsky.social · 20/09/2026

I think the whole “is it really intelligent” question, while interesting, is a red herring, unless we’ve decided it’s actually a being that ought to have rights, which is a whole new can of worms.

100

Hugh Cayless @hcayless.bsky.social · 20/09/2026

I’m not sure they are except in terms of scope. Your coffee maker can’t cause the level of damage that an AI agent potentially could. Presumably you could be liable if your coffee maker burned someone else. I would argue you should be liable if your agent harms someone else. Or ends the world.

100

That would of course entail a level of international cooperation I doubt we’re capable of in this moment.

040

It’s definitely too late if you don’t have the technical expertise and infrastructure on hand or the money to acquire them. We ought to be kicking people off the internet until they patch their shit and arresting the perpetrators but I don’t see any acknowledgement that it’s even a problem.

jamelle @jamellebouie.net · 18/09/2026

Reposted by Hugh Cayless

i am genuinely bewildered by people who say that the rise of white nationalism in the republican party is a response to the "woke left." like, did the woke left cause pat buchanan's 1992 presidential campaign? is the woke left why steve king had a long career in congress?

431103821923

So “scared” is completely the wrong framing. “Furious” is more like it.

170

Anthony Moser @anthonymoser.com · 18/09/2026

We’re not scared. These things produce such a phenomenal amount of traffic that our systems keep going down if we don’t aggressively block bots. It’s pure self defense. And yes, it’s breaking everything.

1184

Reposted by Hugh Cayless

a pretty good way to gauge the conversation is whether people are talking about harm or risk harm is about what's happening now already risk is about what could maybe happen in the future and basically the whole ai cult scene is in agreement that possible risks are more important than current harm

12438143

Um, just, no. 🤦🏻‍♂️

Matt Henry 🥄 @matthenry.fyi · 15/09/2026

Reposted by Hugh Cayless

The US Web Design System is load-bearing infrastructure for thousands of government websites. It makes it easy for small teams with no resources (i.e. most .gov teams) to make usable, accessible websites that work for everyone. Today, it's as close as it's ever been to being replaced with AI slop.

8330141

I also do not enjoy talking to computers, although I do get that it can be necessary from an accessibility standpoint. But I think that’s a different case.

Probably not, as long as you have stuff you are genuinely interested in. You’re not required to be an expert in everything. There’s plenty of stuff I’ve looked at and gone “nah” over the years.

110

Also, there’s a Mississippi River’s worth of terrible programming out there already. AI will add to that, and we’ll have to be much more careful in what we trust, but I’m not sure that’s an insurmountable problem. Less reliance on random packages would be a Good Thing.

020

How much you care is probably what should determine whether you want to use it or not. It can help with stuff you’re not really competent with yet. The question is really whether you actually want to learn how to do the thing. You might well not, and that’s fine.

310

I think in teaching it’s almost always the point. My usual question is whether I have anything I don’t care about enough to use AI on it. So far, the answer is usually that I don’t. I care about communicating clearly and learning stuff 🤷🏻‍♂️

Oh, I’m not saying give up. I’m saying don’t play the game the way they want you to play it.

I’m upsettingly relieved that I don’t have to be in solidarity with Macklemore.

050

Tbh, in my day job, I think they’ve wasted far more hours than they’ve saved. Via different routes.

This is all about making the next round of billion dollar companies, not about improving the human condition.

I suspect it can’t do any good, and you risk doing yourself real harm. Plus, if it really is the key to super productivity, then I can always lessen it later.

My attitude towards AI is probably improperly atavistic, but maybe having conversations with the demon in the box isn’t good for you?

241

Yeah. If there’s any ray of light, it’s that this isn’t sustainable for the scrapers either. They’re not getting anything if they’re just taking sites down. Not sure at what point they realize that, of course.

These things are optimized for biting out huge chunks of, say, the NYT. Not for your cultural heritage site that doesn’t even have a dedicated security team.