iiinova — where your digital self keeps evolving

The problem in these days is less the AI hallucinations, and more the human ones being espoused by these assholes in power.

010

Reposted by Kris Nuttycombe

William Gillis 🏴 @rechelon.bsky.social · 23h

Every decent human being on the internet simultaneously had to say to themselves, "no, wait, not allowed to say that." There's kinship in that experience.

06310

Kris Nuttycombe @nutty.land · 15h

The stories we have of “rogue” AIs breaking out of sandboxes ans so forth all involve the AI being creative in carrying out the instructions of their prompters. Humans are setting the intention, and using directed randomness as an artistic tool.

000

Kris Nuttycombe @nutty.land · 15h

What AI doesn’t have (as far as I can tell) is intentionality. It doesn’t have intrinsic motivation to act, until it is first acted upon.

120

Kris Nuttycombe @nutty.land · 15h

Lots of people in this thread are hung up on questions about “creativity”. But creativity is about making things - look at the word! create-ive AI is good at making things, in a way that previously only people were, and it’s understandable that this will freak out a lot of people.

200

Kris Nuttycombe @nutty.land · 30/09/2026

I had cardamom pastries when I was in Finland and the knowledge of sweet cardamom buns was life-changing.

110

Kris Nuttycombe @nutty.land · 30/09/2026

I’m a geologist, by education. A programmer (and I think a pretty capable one) by trade.

030

Kris Nuttycombe @nutty.land · 29/09/2026

Before I started working from home, my ride was 5 miles. I timed myself on it daily, and timed myself driving on the days I took the car because of bad weather or needing to run other errands. The bike was *always* faster.

011

Kris Nuttycombe @nutty.land · 29/09/2026

In attacking HP's incoherence, it gives readers the opportunity to think to themselves "I thought of that too!" and that draws in readers.

010

Kris Nuttycombe @nutty.land · 29/09/2026

Most sci-fi and fantasy aren't as incoherent as the HP universe, and so wouldn't provide nearly as target-rich environment as HP does. Not only are HP's rules incoherent, they're _obviously_ incoherent even to casual readers. So that's part of the fanfic's appeal.

110

Kris Nuttycombe @nutty.land · 29/09/2026

With all the discourse, I decided to read it and... I'm kind of disappointed in the fact that people are critiquing it on the basis of it being a fanfic. For reaching a broad audience, HP's ludicrously internal inconsistencies make it a perfect _target_ for "rationalization."

120

Kris Nuttycombe @nutty.land · 28/09/2026

The amazing thing is that if these farmers actually pursued this sort of thing, they’d make a lot more money too. And they have plenty of open land to put solar arrays on, too!

1170

Reposted by Kris Nuttycombe

Brosemite Sam @brosemitesam.bsky.social · 27/09/2026

I think we should punish Colorado River valley farmers by building data centers on their land

318115

Reposted by Kris Nuttycombe

rev. howard arson @theophite.bsky.social · 27/09/2026

yes, i am saying this. literally this. i am saying that you should not care about water for 16,000 people when "growing alfalfa for export in deserts with approximately zero precipitation" is literally orders of magnitude more water and vastly more pointless.

3198592

Reposted by Kris Nuttycombe

Julie S. Lalonde @julieslalonde.bsky.social · 25/09/2026

- Social media brain rot where we've failed to see that we are human beings connected to other unique human beings and therefore, just because something doesn't reflect my experience, it doesn't mean it's wrong or "bad" or something you can't still read/watch/play, etc.

21764114

Reposted by Kris Nuttycombe

Dr. Samantha Hancox-Li @sjshancoxli.liberalcurrents.com · 26/09/2026

which is where i will claim my little piece of prescience: woke 2 is woke 1 with an honest relation to power. if the dyke bar doesn't make a profit the dyke bar's going to close. if the standardized test reveals racial disparities, that's a problem with the system, not the test.

liberalcurrents.com

What Is Woke 2?

Woke 2 is Woke 1 with an honest relation to power.

517016

Kris Nuttycombe @nutty.land · 27/09/2026

Suddenly it’s much more clear why NPR has gone to shit.

000

Reposted by Kris Nuttycombe

ellie lockhart (she/they) @eleanor.lockhart.contact · 27/09/2026

Like, every single argument for why markets are bad fails when you consider that the alternative to markets is a system where people are given power instead based on arbitrary sentiments & not required to stake anything, and evaluated on subjective measures

171

Reposted by Kris Nuttycombe

ellie lockhart (she/they) @eleanor.lockhart.contact · 27/09/2026

The choices we have are markets or mob rule, and after decades of rainbow capitalism followed by a crackdown motivated by populist sentiment I find it bizarre how many queer people still want to take their chances with the sentiments of the people over cold hard numbers

1155

Kris Nuttycombe @nutty.land · 24/09/2026

“BCBSA said providers used AI ⁠technology to identify secondary conditions by scanning existing patient records or using ambient scribes” Seems like good news? People are getting treatment they need that might otherwise have been missed?

100

Reposted by Kris Nuttycombe

David Ho @davidho.bsky.social · 24/09/2026

This gatekeeping question is inevitable: "Does the project involve any work that could be interpreted as scientific research?"

856027

Reposted by Kris Nuttycombe

Jason Reifler @jasonreifler.bsky.social · 24/09/2026

While “constitutional crisis” has barely kept any meaning under Trump, this still qualifies — the White House defying a judicial order to respect the 1st Amendment. www.theguardian.com/us-news/live...

theguardian.com

CNN and Politico blocked from White House hours after judge temporarily restores access – live

A federal judge had ruled Trump administration must immediately return the outlets’ press passes and said ban likely unconstitutional

55525

Kris Nuttycombe @nutty.land · 24/09/2026

gdi he read “The Metamorphosis of Prime Intellect” and got fiction confused with reality (again) didn’t he

190

Reposted by Kris Nuttycombe

mcc @dryad.technology · 23/09/2026

Faced with Democrats making historic and unambiguous errors and then losing elections, there's a group of people who decided the problem is people failing to clap hard enough. They then conclude the *specific* people letting everyone down is whoever they personally find most annoying on social media

47510

Kris Nuttycombe @nutty.land · 23/09/2026

Importantly, if you *did* find that you’d made a mistake in turning off an AI you can just… turn it back on.

020

Kris Nuttycombe @nutty.land · 23/09/2026

In particular, the death penalty as punishment doesn’t make any sense, because death can’t reform anybody, and it doesn’t make sense as defense because we don’t need to kill someone to defend society from them.

000

Kris Nuttycombe @nutty.land · 23/09/2026

…but I think it’s still pretty relevant; the whole question of alignment is about how do we protect people from AI and the misuse of AI? “Punishing” an AI doesn’t make sense - I mostly think punishing humans *also* doesn’t make sense, and even for similar reasons.

120

Kris Nuttycombe @nutty.land · 23/09/2026

/justice/ is a little too squishy for me to really say whether I agree here or not; it’s too tangled up with the notion of punishment, whereas I think the protective function is more clear-cut. And I know this is diverging a little from the original thread…

110

Kris Nuttycombe @nutty.land · 23/09/2026

Yup, and importantly that’s not at all incompatible with us having a right to collective self defense that implies some individuals’ rights may be curtailed in service of that defense. The important distinction is that this curtailment should exclusively serve that defense, never punishment.

120

Reposted by Kris Nuttycombe

Teiraa Kaika 🧡🐯🔞 @teiraa-kaika.bsky.social · 22/09/2026

Only people against progress and free information are against libraries, defunds them, shuts them down, bans books. Its easiest to tell who does not have the best intentions for mankind by how they treat books.

1265

Reposted by Kris Nuttycombe

Pookleblinky @pookleblinky.bsky.social · 21/09/2026

It shouldn't be this hard for dems to look at the *wildly* unpopular human centipede of billionaire pedophiles, failsons, rapists, thieves, con artists, and weird online dick pill merchants, all gleefully joining in trying to exterminate trans people, and go "wait a second, let's disagree with them"

121151

Reposted by Kris Nuttycombe

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

This leads to a credibility gap: to put it crudely, why should anyone trust AI safety analysis when y'all are partying with SPECTER and cashing checks from Blofeld? Why should anyone take analysis of lone wolf/PLA misuse of open-weight models seriously when there's no similar analysis of XAI?

626651

Reposted by Kris Nuttycombe

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

The Bond Villain, in this model, is a billionaire who is de facto unconstrained by nation states. They have access to large-scale compute and (via influence or direct ownership) private access to closed weight models that allow them to circumvent any safeguards put on their public use.

216915

Reposted by Kris Nuttycombe

Thomas Dickerson @elfprince13.mumak.app · 21/09/2026

If every two bit script kiddie has access to the same caliber of bugs that were previously only accessible on a state actor budget I bet a lot more of those bugs will get patched 🤭

062

Reposted by Kris Nuttycombe

Kris Nuttycombe @nutty.land · 19/09/2026

My first reaction to Jev was, this is really interesting, and I immediately built it into my dagny.co side project. And this morning it occurred to me that having cheap classifiers built into everything is going to likely bring surveillance hell to every aspect of human life.

dagny.co

Dagny

2163

Kris Nuttycombe @nutty.land · 19/09/2026

dagny.co

Dagny

2163

Reposted by Kris Nuttycombe

Doll @dollspace.gay · 17/09/2026

As always the best way to learn about AI is via weird trans people on bluesky.

0428

Kris Nuttycombe @nutty.land · 18/09/2026

Look, the Rationalist sex cult has a really simple explanation. If you're an insufferable dweeb, nobody's going to pay any attention. But if you're an insufferable dweeb _with a sex cult_, people will think maybe you're kind of cool.

000

Kris Nuttycombe @nutty.land · 17/09/2026

A function is a proof of the theorem that is that function's type. It's a fact about the universe. We build more complex theorems out of simpler ones; the top-level theorem (articulated as a spec) is simply that, the top level, not all the layers below it.

000

Kris Nuttycombe @nutty.land · 17/09/2026

Software is specification, but precise. It's mostly that the languages we write software in suck. The way that they suck is that they flatten complexity at different levels of abstraction.

100

Reposted by Kris Nuttycombe

Paul Crider @paulcrider.liberalcurrents.com · 17/09/2026

I just think, of all the crazy ideas campaigns we've seen in this country that have succeeded or come close to succeeding, open borders isn't obviously nuts. Abolition, women's suffrage, gay marriage, the Civil Rights movement, ending the war on drugs. All of these had massive headwinds.

021

Kris Nuttycombe @nutty.land · 17/09/2026

I should really make time for in-person go. For whatever reason, I find online go incredibly anxiety-inducing; got to have the tactile beauty of the board and the stones to compensate for the stress of the gameplay.

130

Kris Nuttycombe @nutty.land · 17/09/2026

20 y/o me is very much "That's not at all what I expected, but ok, I'm kind of impressed."

040

Kris Nuttycombe @nutty.land · 17/09/2026

Also, AI code was uniformly slop a year ago. That has changed so much in the past year that if I hadn't experienced it directly, I would simply not believe that this much change was possible.

020

Kris Nuttycombe @nutty.land · 17/09/2026

Even if all they were doing was bug-hunting and cleaning up known technical debt, they'd be getting effectively a doubling of their workforce for a 10% increase in cost.

100

Kris Nuttycombe @nutty.land · 17/09/2026

That little? Theyt should be spending that much on AI-assisted bug finding, just to start with.

110

Reposted by Kris Nuttycombe

Technology Connections @techconnectify.bsky.social · 17/09/2026

As the opposite of a computer toucher, I sometimes worry that I'm not keeping up with the development of things. Luckily, I'm smart enough to realize that if the "most dangerous" LLMs need warehouses full of GPUs to function, they can't just "break out" and start running on random PCs.

631301103