Jonathan Cheng

💯 my model for who is looking for market/ legal remedies vs federal/ social remedies is very unreliable these days

Thinking: “too many model releases/ announcements today, ugh.”

*looks at job title*

Damn.

openai (OpenAI) Org profile for OpenAI on Hugging Face, the AI community building the future.

OpenAI is back with open source releases on @hf.co. This is the biggest release I've seen in a long time!

See more in huggingface.co/openai

Told the bf that liking “Soda Pop” more than “Golden” or “What it Sounds Like” is a relationship crime.

15-20 years ago, I’m imagining trying to convince record labels that the top 10 would be dominated by the kpop ost, of an animated show, on a streaming service.

Oh, for sure!

If I felt that the vibes were distributed between distanced critique and participatory good-faith explorations, then there wouldn’t be much to say — but, to my surprise, more folks on bsky have been popping up and saying that is definitively not the env we’re in.

That any technology might lower the barriers of entry to exploring different disciplines?

Sounds suspiciously like liberal arts and sciences to me!

A lot of times folks on the left adopt the logic they critique, happily clapping along when there are signs that LLMs aren’t immediately providing hyped amounts of utility.

There’s a lot to say about that. But I’m already excited by what I’m seeing. Ie. Artsy folks vibe coding indie games.

Terence Tao has written a thread about the damage being caused by the grant freezes at UCLA.

mathstodon.xyz/@tao/1149568...

I am delighted to share our new #PNAS paper, with @grvkamath.bsky.social @msonderegger.bsky.social and @sivareddyg.bsky.social, on whether age matters for the adoption of new meanings. That is, as words change meaning, does the rate of adoption vary across generations? www.pnas.org/doi/epdf/10....

In some strange zeitgeist/ instagram influenced moment, I was just looking this up 😂

alignment.anthropic.com/2025/sublimi...

Oh interesting, student models/ fine tunes might pick up statistical signatures of a parent model’s internal state.

…But can Anthropic convince me that this wasn’t written by owls? I think not.

Oh yeah, all the new agent based language + the ml language 😂

Cog Sci students who specialize in ml are extremely well positioned — everyone else tho… 😂

And I’d argue that controlling the latter three components is a losing battle.

So I feel like the only place to get more traction is to make sure college students take certain courses.

💯

The combination of educational prerequisites + reactionary social media accounts + large vocab of tired sci fi tropes + fast rate of change?

It becomes very hard to have a cogent conversation about the topic.

Lolol, “how long of a reasoning inference run = agency?”

This came up because someone took back their surprise about Claude solving a math problem, being like “actually, I think it could do this already.”

And, actually, a year ago, no it really could not have.

The pace of llm developments is so fast that it makes it easy for humans to hallucinate that nothing is changing/ improving.

Ie. Folks will sometimes say “well LLMs were always going to be good at math and code” — when like a year ago LLMs doing basic arithmetic was the topic of conversation.

Congrats!!!

That makes a lot of sense to me!

But it’s definitely possible I should just get used to navigating multiple code agents at once — I might just have some unfamiliarity based reservations! 😂

That’s my impression, and I worry about collisions between multiple agents handling my shorter term queries (maybe unfounded).

Maybe I want something like a short term agent that handles my “implement this” queries — and a longer term agent that only reorgs stuff I’ve written.

One part of the llm-coding experience that I’m not a huge fan of is the downtime between queries.

Would be cool if there was a tool that was like “other design decisions/logic that you’ll need to think about.”

Kind of like anticipating stuff I might need to consider, design patterns, etc.

Actually, now that I think about it, it fits into a nice history of “the gothic and aesthetic critiques of real estate” — capped off with Paranormal Activity and the McMansion.

I think that was my experience.

I only happened to see a link to The House on the Moor, thinking that the title was hilariously on the nose, and I was happily given exactly what was advertised — plus some occasional lines of interior decor critique 😂

Charles Chesnutt, Margaret Oliphant

I think I was supposed to pick 1 😂

I felt like I learned about George MacDonald and George Gissing pretty late? I guess this is defining underrated as under-taught, less an aesthetic judgement about their writing itself.

100% — alas, we don’t always say these things in a progressive/ city-planning sort of way, as we sometimes say it in more of an eff all motorists kind of way 😂

My partner and I are both avid bikers/pedestrians who avoid narrow death — not infrequently.

And I think it has caused us to mutter uncharacteristic things like:

“Car owners are fundamentally broken.”
“When I’m mayor, I say we ban all cars.”

Especially relative to any of the other options 😂