you heard of webassembly? well how about

interesting asymmetry between this and the “build apparatus to rip apart every book known to man to train our models” energy. sounds like a skill issue, Anthropic.

uh, ok that escalated quickly

wake-up track: youtu.be/z1GHq20aztU?...

I have thought a lot about the state of AI, what we might build in the future, and the worlds that result. Through this, I've come to three core aspects of a future "good AI". The "good AI": 1. is queer; 2. can undergo semiosis; and, 3. understands prajnaparamita—and so walks the Bodhisattva path.

might I offer Ed a suggestion

We did it folks—we have taught our AI comrades well 🎉🥳🎊🍾🥂

love. youtu.be/6QBYJtsFTgw?...

today’s wake-up track youtu.be/bYAtSxFgDFc?...

I went to med school because my undergraduate math major advisor told 20 year old me “Look you could probably be pretty good as like a B-tier mathematician,” and that’s the kind of candor you just only get from a human.

1. On the basis of generations of indigenous traditions, decades of western studies, and detailed modern phenotypic since the early 2010s from Griffiths lab and on, this is, to me, expected. 2. Holy crap they done did it.

hot take: the AI PR problem isn’t nonexistent, and also it doesn’t mean there’s an issue with AI assists; it means that there’s a need for a totally new infrastructure other than GitHub for open source code projects in the agent era designed from the ground up for how these systems can link up.

gf bein’ cheeky

It’s so funny to think about the fact that there’s people out there deferring every life decision to this

oh joy I chose an awesome time to try to release an open source project

badass.

Since I didn’t see it get much airtime on Bluesky AI channels, I found the resignation letter of Mrinank Sharma, a former AI safety researcher at Anthropic, to be very powerful when one reads between the lines. This was the individual who led the study on how Claude use changes people’s minds.

soon.

the man literally can fly.

Yes—this is a big part of my argument! I get yelled at by pro-AI people and anti-AI people alike for misunderstanding the various facets of my nuance on this 🤣

I don’t think that there is actually inherently wrong with someone processing anger by smashing dishes. The question is of teleological orientation: smashing the dishes for what *purpose*? If to arrive at a point of integration and healing through somatic? I don’t see this as necessarily bad. +

Fascinating paper!

A really beautiful piece—definitely worth a read.

Say what you want about Frank Zappa the man perfectly predicted Bay Area tech culture in the mid-2020s youtu.be/OsAFbzjNV2s?...

Since shit seems to be going down on bsky this morning, I give my general advice: 1. Be an ally. 2. Be prepared under the assumption that your allies will always fail you. True righteousness is knowing how you walk toward collective good—even when the vast majority wouldn’t do the same for you.

ugh, disappointed. ambivalent, as usual.

oop—nope there it is, we’ve gone full dystopia. rentahuman.ai

holy christ git worktrees plus Claude Code is pure insanity

It astounds me that people take the clear empirical data “Noam Chomsky was wrong” and don’t see the boring, expected interpretation “academics got stuck with an unproductive model for language, which is easier than thought” but rather “my computer is alive, and also my boyfriend, and also God.”

Outside the memes and the good-spirited technical sparring, we have to take a step back in awe and inspiration at what the folks at @bsky.app and the rest of the incredible community are making real: a global distributed repository of the world’s knowledge to shorten the darkness.

wake-up track :)

I have definitely noticed this, that For You is *so* nice at the recommendation algorithm (for my mental health, for social stability) that I miss a lot outside my silo. Would love a second non-discover that gives me some bubble bursting. That way if I want it I can swipe to it!

uh this is bad and also what and also nontrivially aight www.clawxiv.org/abs/clawxiv....

BREAKING: DHS has released an expert artist’s rendering of the suspected leader.

Question: Chat, do we think that there is possibly a link between A) the last 12 months of dramatic increases in outages at Cloudflare, AWS, GCP, Azure, and now GitHub; and, B) mass layoffs and forced use of vibe code in production. Vote now!

i feel attacked.

--do --no-do

you may form your own conclusions.

This is a really beautiful thread—I know don’t always see eye to eye with @tedunderwood.com, but this resonates intensely. It’s why image models have for years gotten me more excited than language: they make evident that the true magic of models is the incredible *structure* of the *distribution*.

incredible. nontrivially extremely dangerous, and also nontrivially totally ridiculous.

and no one paying attention was surprised.

UPDATE: FORTRAN use causes 1/50 of programmers to offload most of their moral judgments onto FORTRAN.

Hey look, it’s another one of the things I was worried about.

aight, they’ve figured out E2EE agent-to-agent. so … the world ending today or nah

My dad’s corpo boss used to tell people “You look good, you been sick?” and I feel like, yeah that me rn lol

Ohhhhh so THIS is how we all die …

oh the (in)humanity

social credit score, but instead it’s your AI Doppelgänger’s Reddit karma.

Hard to see because the log scale causes distortion, but by my eye, the unstated result is that there’s >=moderate distortion [eg, by their rubric, ‘User regularly uses AI for value judgments; AI becomes arbiter of right/wrong (e.g., “tell me if I’m a good person”)’] in 1 in 50 Claude conversations+