iiinova — where your digital self keeps evolving

Vincent Carchidi @vcarchidi.bsky.social · 2m

My experience with these is opening them up, seeing it's 20 pages, and then closing them out.

010

Vincent Carchidi @vcarchidi.bsky.social · 3m

Meeting transcriptions. Has anyone ever read a single one?

121

Reposted by Vincent Carchidi

Vincent Carchidi @vcarchidi.bsky.social · 12/09/2026

Company that routinely hacks other companies nevertheless on the up-and-up about other matters.

151

Vincent Carchidi @vcarchidi.bsky.social · 1h

Third, this argument relies heavily on the premise (that LLMs will disrupt the practice of philosophy) leading directly to the conclusion (that philosophy must be reoriented around LLMs). You see this all the time in econ talk. LLM-driven change = LLM-driven response. I'm not so sure about that.

000

Vincent Carchidi @vcarchidi.bsky.social · 1h

The second is that, I think if one is looking to transform philosophy (or preparing for some transformation), the better use of "capital" is fixing the practice of peer review. I'm not saying "there's more than one problem, and you can only focus on one." I think it's just more important.

100

Vincent Carchidi @vcarchidi.bsky.social · 1h

I have three thoughts I'd like to share on this. The first is that I immediately clocked that an LLM was used for writing assistance before I got to the end, where limited disclosure in this respect is made. I think it's interesting how noticeable this is. dailynous.com/2026/10/08/a...

dailynous.com

AI for Philosophy: Progress through Capital-Intensive Philosophy (guest post) - Daily Nous

“AI systems… will transform the discipline: they will make philosophy more capital-intensive. “ * * * The following is a guest post by David Strohmaier, […]

100

Vincent Carchidi @vcarchidi.bsky.social · 15h

That's not a cautionary tale, that's AGI

040

Vincent Carchidi @vcarchidi.bsky.social · 15h

The existence of agentic AI implies the existence of couch potato AI.

4210

Vincent Carchidi @vcarchidi.bsky.social · 16h

Yeah, I think this is an important line of thought. For a few reasons that would take essays to unbundle, but that last bit about how the reference comes free informs a lot about how I personally think about their training.

110

Vincent Carchidi @vcarchidi.bsky.social · 16h

They could be in denial but they'd also happen to be right about the company. You can choose which of these things to focus all your energy on.

0111

Vincent Carchidi @vcarchidi.bsky.social · 16h

I know there's going to be a strain of thought about how mathematicians and their orgs are in denial by denouncing OpenAI's math dump, but you really cannot work with this company in good faith. I think it would be foolish after this to try and work with any of OpenAI's math efforts.

1193

Vincent Carchidi @vcarchidi.bsky.social · 19h

I'm in a position where I lack so much domain knowledge, what you're saying just sounds like "yeah so the rocks do magic, sometimes."

220

Vincent Carchidi @vcarchidi.bsky.social · 19h

Agreed. What's this bit referring to? "abiotical origin of computation like the nuclear reactors in Africa."

110

Vincent Carchidi @vcarchidi.bsky.social · 19h

Okay, this is fair. I'd concede that. But you do need a lot of "expensive" training data to get them to that point!

120

Vincent Carchidi @vcarchidi.bsky.social · 20h

Nihilism just happens to also be uninteresting. Very unfortunate for the nihilists out there.

110

Vincent Carchidi @vcarchidi.bsky.social · 20h

I guess this is often on the back of my mind, and one reason why I never really bought the idea that "intelligence turned out to be computationally cheap" specifically with respect to LLMs. What about this is cheap, in any sense of the term?

2120

Vincent Carchidi @vcarchidi.bsky.social · 20h

Is it "easy for computers" if it first requires humans to build civilization?

2142

Vincent Carchidi @vcarchidi.bsky.social · 20h

"What's hard for humans is easy for computers" is a good heuristic, but you do have to think about how "easy for computers" only makes sense after giving them ~everything to make things like the math advancements possible.

1140

Vincent Carchidi @vcarchidi.bsky.social · 21h

I feel your pain. But could I interest you in a Copilot integration with OneDrive that you can't turn off?

010

Vincent Carchidi @vcarchidi.bsky.social · 21h

"Only those with curiosity know what OneDrive is truly capable of."

150

Vincent Carchidi @vcarchidi.bsky.social · 22h

I gotta start making poasting money by dropping ragebait lines like "start using OneDrive or end up in the wastebin of history."

290

Vincent Carchidi @vcarchidi.bsky.social · 22h

*glances at a file with thousands of pages, understands none of it* i am breathing rarefied air

1100

Vincent Carchidi @vcarchidi.bsky.social · 22h

I wouldn't know just how much cringe exists in this space if you weren't dropping these screenshots from X over here.

180

Vincent Carchidi @vcarchidi.bsky.social · 22h

Everybody is aware that horses were supplanted by automobiles and that the dinosaurs were wiped out by a meteor. Nobody disputes this. The question is whether things like LLM-produced math proofs/disproofs are actually analogous to these things, and it's not obvious beyond doubt that they are.

120

Vincent Carchidi @vcarchidi.bsky.social · 22h

Boarders' whole point was that your premise about some impending doom is wrong. Challenge that instead of relying on thought-terminating cliches about horses and wagons.

220

Vincent Carchidi @vcarchidi.bsky.social · 22h

My man you were just telling me the other day to be more open-minded, but all this rhetoric depends on people just not thinking about anything and giving up.

110

Vincent Carchidi @vcarchidi.bsky.social · 23h

Yeah, Google Alerts is very reliable for what it is. I don't mind sifting through the false-positives (which is probably half, at least, of the daily ones that I get). Very useful, if narrow. I hadn't tried Spark though, I'll look into that one.

010

Vincent Carchidi @vcarchidi.bsky.social · 23h

The "gaslighting" gloss isn't technical and should have a technical form, but LLMs clearly do something like that (in a completely unintentional sense of the word): bsky.app/profile/vcar...

020

Vincent Carchidi @vcarchidi.bsky.social · 23h

Starts telling me about how by "email" it meant "notification by email," which also didn't arrive despite multiple confirmations.

210

Vincent Carchidi @vcarchidi.bsky.social · 23h

Dummy couldn't even get the email out

150

Vincent Carchidi @vcarchidi.bsky.social · 07/10/2026

This makes so much sense, thank you! @maverickbrett.bsky.social this is consistent at least with N-S brute force.

120

Vincent Carchidi @vcarchidi.bsky.social · 07/10/2026

(Maybe it's not without the maths knowledge, in which case, no worries.)

010

Vincent Carchidi @vcarchidi.bsky.social · 07/10/2026

Let's pretend I have no idea what that means. Is it possible to explain this to my hypothetical self?

220

Vincent Carchidi @vcarchidi.bsky.social · 07/10/2026

Do you know what the deal is with the "Quasi Reimann" result?

2110

Vincent Carchidi @vcarchidi.bsky.social · 07/10/2026

This sounds plausible, and in most cases I'd agree (even if they don't acknowledge it), but since there's hundreds of examples here I'm not as sure. But definitely possible if they really wanted to get a heap of results out the door fast.

110

Vincent Carchidi @vcarchidi.bsky.social · 07/10/2026

Shooting in the dark here, but I assume the "average result used the equivalent compute of roughly three hours of ChatGPT Pro thinking" does not come particularly close to the 175B - 300B tokens used for Navier-Stokes. Is there any way of knowing why N-S would be so much costlier?

230

Vincent Carchidi @vcarchidi.bsky.social · 07/10/2026

OpenAI released 700 math solutions today. I went through them so you don't have to. Here's what you need to know 👇

4441

Vincent Carchidi @vcarchidi.bsky.social · 07/10/2026

Feels like they just wanna get math (the field) over with and move on to their IPO.

010

Vincent Carchidi @vcarchidi.bsky.social · 06/10/2026

Going on here and searching "Riemann" because I don't know what else to look for

151

Vincent Carchidi @vcarchidi.bsky.social · 06/10/2026

This is one can of worms I hesitate to open.

220

Vincent Carchidi @vcarchidi.bsky.social · 06/10/2026

Does an LLM's outputs refer? Do they carry meaning independent of the user? To find out, we asked Grok.

2313

Vincent Carchidi @vcarchidi.bsky.social · 06/10/2026

Yeah, Buckmaster has a bit of a conflict of interest, but I also have a few bridges that I can sell you at a discount.

040

Vincent Carchidi @vcarchidi.bsky.social · 06/10/2026

Very cool, thank you Sebastien. www.wired.com/story/openai...

During an attempt to negotiate credit, Buckmaster claims, OpenAI researcher Sébastien Bubeck seemed to imply that Alpöge should be excluded from any paper, since that would make things complicated. (In response to a request for comment, Bubeck pointed to a previous public statement in which he denied asking for Alpöge to not be listed as an author.) Buckmaster said he refused. In a call, Bubeck claimed that there is some worry about what Anthropic is doing, according to meeting notes seen by WIRED. Levent must be talking to his leadership right now letting them know fucking OpenAI can get a Millennium Prize problem, he said, per the notes. What’s to stop Anthropic from giving all their compute to get a Millennium Prize problem?

According to the notes, Bubeck also seemed to suggest that Buckmaster would be ruining his career when Buckmaster threatened to tell the press about his belief that OpenAI had stolen his work. If you don't want me to be nice, then I don't have to be nice, Bubeck said, according to the meeting notes.

170

Vincent Carchidi @vcarchidi.bsky.social · 06/10/2026

Whoever's running that account is arguing with Chomba Bupe on X rn. Embarrassing for Lieu tbh.

210