iiinova — where your digital self keeps evolving

Human-authored or human-prompted? I think it's likely that a developer explicitly asked Claude or ChatGPT for "a Minecraft Easter Egg" and snuck it in. But I would not really read anything into the particulars of the text other than "someone likes Minecraft and Easter Eggs".

020

Reposted by Andrew F

Meredith Rose @mrose.ink · 29/09/2026

In short: Revive the OTA. 11/11

6778

Andrew F @aftd.dev · 29/09/2026

Fuck, we've been paperclip-maxxing this entire time.

010

Andrew F @aftd.dev · 29/09/2026

It is kinda funny that current military trends suggest that the future of war is less guns and more "magic arrow chases you down".

static.klipy.com

Rocket Raccoon and Yondu Fighting

Alt: Rocket Raccoon and Yondu Fighting. Rocket is shooting guns and Yondu has an arrow chasing people down

000

Andrew F @aftd.dev · 29/09/2026

Well, there's a whiter side.

020

Andrew F @aftd.dev · 28/09/2026

Moravec's paradox but for x-risk

010

Reposted by Andrew F

Red Sky At Night @sailorsdelight.bsky.social · 24/09/2026

We must pause agentic children.

1172

Reposted by Andrew F

Faine Greenwood @faineg.com · 26/09/2026

like I suspect many of you on This Website, I think a core reason why I find Eliezer Yudkowsky and his rationalist and effective altruist satellites so intensely annoying is because I had a VERY similar “being a super weird, pretentious, hyper-literate kid” experience - I just, you know, grew up.

532615263

Andrew F @aftd.dev · 26/09/2026

My instinct these days is that LLM are intelligent in the way that slime molds or fungi might be intelligent. And that's kinda cool! Maybe even useful. But I have to admit that the personhood of slime molds is not a Top 10 ethical quandary for me.

010

Andrew F @aftd.dev · 25/09/2026

I work with a Phillipines team, and there's something to that ambivalence at the bottom of this chart. They've been much more aggressive about using AI to paper over linguistic / knowledge / timezone / context gaps than their US counterparts.

010

Reposted by Andrew F

Bert Spaan @bertspaan.nl · 24/09/2026

I've made an interactive map for exploring the terrain of the world! It uses @mapterhorn.com data and tiles, and WebGL for dynamic colors and animations. relief.bertspaan.nl/world?lang=en

35110

Andrew F @aftd.dev · 24/09/2026

> if the defendant possesses anything that is inherently dangerous, as specified under the "ultrahazardous" definition, the defendant is then strictly liable for any damages caused by such possession I dunno, maybe we just take them at their word that LLMs are super dangerous.

en.wikipedia.org

Strict liability - Wikipedia

000

Andrew F @aftd.dev · 22/09/2026

I wonder how much student AI writing could be "detected" by simply printing out a random page from each student's essay, removing formatting and identifying marks, and then asking each student to pick out the one that was theirs.

010

Andrew F @aftd.dev · 21/09/2026

Does this actually protect bsky though? One way to read this is that bsky the social network is not responsible for the content of user selected feeds. But bsky is also a feed provider, and in that capacity, it might still be on the hook?

000

Andrew F @aftd.dev · 21/09/2026

Oh yes, 100%. bsky.app/profile/aftd...

010

Andrew F @aftd.dev · 19/09/2026

Is tree removal spam suddenly a thing for anyone else? I get a ton of these "we're doing tree removal in your area" messages but they're from different, not local numbers and they never mention a name.

000

Andrew F @aftd.dev · 19/09/2026

"There is no point in banning guns because then people will use knives."

000

Andrew F @aftd.dev · 19/09/2026

So what I'm hearing from you is that data centers make beer more expensive.

000

Reposted by Andrew F

Bay Area News @bayarea.bsky.social · 18/09/2026

Last call: This historic, pioneering Bay Area brewpub to close its doors after 40 years

ktvu.com

Last call: This historic, pioneering Bay Area brewpub to close its doors after 40 years

011

Andrew F @aftd.dev · 15/09/2026

I've developed a newfound respect for Strunk & White's Elements of Style.

010

Andrew F @aftd.dev · 14/09/2026

The honest truth: Sucks that it ruined "honest"

010

Reposted by Andrew F

Oaf @languidoaf.bsky.social · 14/09/2026

CEO of Taco Bell: we’re actually worried that some of our upcoming items could destroy humanity as well

362658515

Andrew F @aftd.dev · 13/09/2026

I don't know, but I know Graham developed an unusually elevated opinion of Altman as early as 2008.

030

Andrew F @aftd.dev · 13/09/2026

Paul Graham liked him and designated Altman as his chosen successor at YC. That gave him a lot of cards to play.

191

Andrew F @aftd.dev · 13/09/2026

"Wait, take a photo first"

0791

Andrew F @aftd.dev · 12/09/2026

"We need to pause AI development" "Aight, let's unplug US-East-1" "No not like that"

010

Andrew F @aftd.dev · 12/09/2026

This is sort of similar to the discussion going on with software devs, yeah? The process is not simply about maximizing feature throughput but about building shared understanding, mentoring others, etc. Likewise, there's more to math than maximizing proof production.

010

Andrew F @aftd.dev · 11/09/2026

*humming "Goooooddddd, what have you done?" as the pole topples*

010

Andrew F @aftd.dev · 11/09/2026

As someone who gives LA shit for car culture, this is because we refer to everything south of Santa Barbara and north of San Diego as "LA"

111

Andrew F @aftd.dev · 10/09/2026

Have any of the AI note taker lawsuits reached some sort of resolution yet? It seemed like some of them were going to try the "it's a summary not a recording" argument which ... dubious, but someone's going to shoot that shot.

100

Andrew F @aftd.dev · 09/09/2026

To date, I still haven't read a good explanation of "it's important we get there first because we're the only ones we trust to act responsibly". Like, getting nukes first does not enable you to deny other people nukes unless, well, is that what we're suggesting?

110

Reposted by Andrew F

Nick Semper Tyrannis @nicksempertyrannis.bsky.social · 07/09/2026

I've done this and the wax falls off instead of pooling and getting carried up the wick and the entire wick burns up in like 20 minutes and at some point the off-balance candle falls off and rolls away while still on fire. I do not recommend.

1112

Andrew F @aftd.dev · 08/09/2026

That said, if I had to pick an example of where Vime's Boots Theory seems pretty true these days, I'd probably pick something like healthcare instead of consumer goods? Not that poor people spend more on healthcare but preventative care pays dividends well beyond the actual cost of healthcare.

010

Andrew F @aftd.dev · 08/09/2026

*Stares in Silicon Valley*

051

Andrew F @aftd.dev · 07/09/2026

OK, but what if public transit game where you stack Mahjong-like tiles to represent population density?

010

Andrew F @aftd.dev · 05/09/2026

Hot take: Follow feed by default removes one of the levers you'd use to mitigate dog piling. "A lot of people are angrily QTing this post so we are going to just stop showing those" is sometimes a sensible approach.

000

Andrew F @aftd.dev · 05/09/2026

Toss up between pretending to be a sexy woman and pretending to be a refugee. Also, weird how both scenarios provide plausible explanations for why you prefer payment in crypto.

000

Andrew F @aftd.dev · 04/09/2026

The Final Destination series has ... uh ... a lot.

020

Andrew F @aftd.dev · 04/09/2026

You know someone at Fox is dying to put up a "Sharia Socialist Sex Symbol" chyron.

020

Andrew F @aftd.dev · 04/09/2026

I sometimes wonder if Claude keeps saying "honestly" because Anthropic decided to turn up the "let's be honest" dial in training, but now we just get a Claude that honestly makes stuff up.

020

Andrew F @aftd.dev · 31/08/2026

There's also the forcing function: you need RTL support when you localize Arabic or Hebrew. Those languages are consistently RTL. Traditional CJK uses vertical writing modes but modern usage embraces horizontal as well, so it doesn't force direction switch in the same way.

000

Andrew F @aftd.dev · 31/08/2026

I mean, how often do you have CSS that has to support both horizontal and vertical writing modes? Seems rare. On the other hand, margin-inline-start instead of margin-left is simpler because LTR and RTL interfaces can be, well, mirror images of each other.

100

Andrew F @aftd.dev · 27/08/2026

Princeton also has courses for "The Age of Dinosaurs" and "The Warrior Culture of Japan", and it's kinda interesting folks don't insinuate those courses are frivolous and useless!

000

Andrew F @aftd.dev · 27/08/2026

LinkedIn posting unfortunately appeals to two kinds of lawyers: • Lawyers who like to hear themselves talk • Lawyers who don't know why that other lawyer is doing a thing but it might be important so let's copy, just in case

020

Andrew F @aftd.dev · 26/08/2026

...wait, is there a way to efficiently simulate traffic using inference? Because that sounds utterly bonkers to me but also if someone published a paper tmrw about how diffusion models are actually good at simulating the interactions between millions of cars, I'd be like, huh, nifty.

000

Andrew F @aftd.dev · 26/08/2026

It's going to be real funny if games in 2030 are doing crazy things with inference, but a moderately sized Cities: Skylines 3 city still manages to slow your machine to a crawl.

130

Andrew F @aftd.dev · 24/08/2026

"Only show people from these PDSes" makes more sense as a custom feed. Which incidentally brings up my gripe that *feeds* aren't very composable. Imagine if ordinary users could create their own algo that weighs a PDS up and certain labels down from directly in the app.

000