I'm working on a paper trying to see if generative linguistics and usage-based linguistics can be unified (from a very specific angle), and so far my feeling is: they can provide inspiration for one another, but it would remove core tenets from either to try and make unification happen.

Hello, new followers. Please be aware my dominant posting style evokes the thought: "why would he post that?"

Read the article before whipping people into a moral panic challenge (impossible)

My vision of the future? It's simple: I think events will play out in ways that conform with my beliefs.

I feel like I should put this out there: if for some reason I don't respond to a question or something, it's probably because it has a word I muted in it and I just didn't see your post.

I'm not sure I buy the whole recursive self-improvement thing even on a conceptual level, and I am even more skeptical of this with currently existing systems.

Stories like this (the stonewalled engagement with the medical field) are just shockingly common in the US...

New blogpost: We Don't Understand Neural Networks At The Algorithmic Level kripken.github.io/blog/neurosc... So, why is this even a question? Don't scientists agree on whether we understand LLMs?

Interesting position. Very reasonable IMO. www.linkedin.com/feed/update/...

Mood

Yeah, I co-exist with AIs Absolute Imbeciles

That's Dario Amodei's brother on the left btw

I think there is an intuition behind the negative reaction consumers tend to have when learning that an LLM was used by the creator which is at least worth thinking about.

While I'm on the topic, I do think this is an area where journalists have completely dropped the ball.

This is a fair reading of the passage below, but IMHO I don't think that's what it's saying.

My ancestors survived by ignoring things like cockroaches so now I gotta get the heebie jeebies every time I see one of these fuckers. Extremely unfair.

Brb making sweeping assessments of the internal states of Google, Anthropic, and OpenAI on the basis of turnover in a high turnover industry.

So, if someone accepts that a skilled user is and will likely remain* necessary to get real use of an LLM in a specific domain, I do not fully understand what the actual pickle is with educating students and so forth. (*The most bullish people probably don't, so this is not relevant to them.)

"We find that humans enrich logical reasoning through pragmatic inferences across languages. Model behavior is more variable." arxiv.org/abs/2605.21299

Is the US preparing a shakedown or did somebody actually managed to smuggle a long truck-sized machine to China? Who knows! www.bloomberg.com/news/article...

Sharing without comment. Looks interesting link.springer.com/article/10.1...

By far, my favorite philosopher is Hofstadter

Interesting!

In fairness to the various Finance Bros, it seems true that, under enough pressure about a looming recession (or worse), the Trump admin. agreed to basically whatever Iran was demanding. On the flipside, I think what's happened since Feb. is not being priced-in properly.

So is this Iran deal actually happening or what

Architect of the Trump administration's AI Action Plan is joining OpenAI.

Didn't Google lock this guy into a contract worth like $2 billion?

Ok foreignpolicy.com/2026/06/17/t...

Re: the idea that leaders at AI labs - or really just Anthropic - believe what they say about RSI, ASI, and so forth. I've found it useful to think about this sort of thing as, yes, they believe what they say, but sincere belief and self-interest can get tied up with one another seamlessly.

An urge to write but no clue to what write about. Awful.

Getting reports that people are doing things with large language models

Preview, not Fable www.nextgov.com/cybersecurit...

I'm building an AI-powered service that alerts you when new AI-powered products you might like become available. Powered by AI.

Sharing not because I'm in tech but because idk what the hell some people in tech are thinking right now.

Currently hearing things about vagueposting 👀

I think the best indirect argument suggesting that LLMs do not do something akin to human reasoning (which is not the same as saying they don't reason at all) is that their writing is not good.

Interesting read. Not sure Carlini is the guy you'd wanna send to ease the government's concerns though. www.wsj.com/tech/ai/anth...

Doesn't make it remotely better, but this is from 2025.

- AI company claims later models will have capability X. - People argue over whether later models will have X. - Company releases model, claims it has X. - Questionable that it has X. - Some people believe it has X, others do not. - Time passes, company claims later models will have Y.

Anybody got any book recs? If we're moots then we're probably interested in similar things anyway, so feel free to just name whatever you might be reading atm.

This spiritually contradicts the Semafor reporting that the reason for the directive was because a China-linked group had *already* accessed Fable/Mythos. www.reuters.com/technology/a...

France is pretty serious about both decoupling the tech it uses for government operations from US firms while also developing a sort of Maven-esque C2 system without Palantir. www.defensenews.com/global/europ...

People are always coming up to me and saying, "your poasts changed my life." I always tell them the same thing: "I know."

One day they're gonna digitally resurrect Al Pacino and call him Al Pacino

I am very interested in this man's arc www.youtube.com/watch?v=vS-g...

All of Western philosophy consists of a series of DLC expansions to Plato.

I will no longer poast about controversial things. It's time for a change. From now on, I will post only about things that there is consensus on. We can all agree that Searle's Chinese Room Argument correctly points out that Large Language Models don't understand a word they're saying.

If there's anything I'm taking from this breaking containment, it's that I'm living with a much different view of human beings than a lot of you. Which is well and good. But interesting to be aware of it.

Updated version of a paper I enjoyed in its ArXiv form: jlm.ipipan.waw.pl/index.php/JL...