love when physicists work it, write some things down, flip it and reverse it youtu.be/x4TdColoIu8?...

I LOVE that Pessimist's Archive has woken up again It explores the history of rejecting technological progress and it is ALWAYS eye opening bsky.app/profile/pess...

I know he wasnt perfect, but he was the best President of my lifetime The dude had NO scandals, he got his healthcare bills passed, AND he got bin Laden bsky.app/profile/acyn...

I love what Ai2 is doing so much allenai.org/blog/domyn-a...

every now and then i see a video critique of society that hits so hard i forget how to speak

autocorrect can go to he'll

Using LLMs to code *feels* more like playing an instrument than it *feels* like writing When I play drums, all 4 limbs are doing their own thing. When I play guitar, I'm thinking about both hands and each finger What happens after that comes from the machinery

Noam Brown is a fascinating thinker First heard of him when this talk was posted online. My inner engineer was still dominant in my thoughts because I kept thinking, "so what? we have algorithms to do all of this" and then I realized he was doing the same stuff via ML instead youtu.be/eaAonE58sLU

Yup! AI is just software, after all bsky.app/profile/mims...

@timkellogg.me is your event based agent anywhere? does strix gave your recent work?

noise is just chill chaos true for noise bands too

Cognitive biases are so incredibly interesting Once you understand what they are and how they work, you can start to understand where your own are BUT! It is very often the case that you cannot see your own biases, because that's how biases work. You *actually* dont know theyre there

mamdani does not miss bsky.app/profile/wuta...

Yup. www.axios.com/2026/06/15/a...

Something I think about a lot The techbros knew AI had a big moment coming soon. OpenAI was founded in 2015, but I remember Altman publicly saying he thought AI was the future before then Soon after, MAGA + techbros began filling the Internet (big chunk of training data) with the worst content

@jbf1755.bsky.social @hcrichardson.bsky.social @paxll.bsky.social ugh have yall seen this..?! bsky.app/profile/aaro...

I wish we could fold threads in bluesky similar to what happens on Reddit

opus 4.7 > 4.8

Im gonna be quieter on this thing. My follower count shot up recently and I got excited, but I feel weird saying so much.

Cant wait to hear NYC go nuts in a few minutes

he's wadario now

What if Trump suspended Fable because he wants everyone to watch his UFC thing

"the usgov has the right to suspend models" models are speech. they cannot suspend them

When I think about the datacenter spending, I worry it's very hard to stop investors once they get going Then I see people yelling loudly enough to force politicians to fight the buildout Antidatacenter folks *could* get every detail about tech wrong, Id still appreciate reducing risk of a bubble

Congratulations to Elon on becoming the world’s first trillionaire from London 🔥

This is a fantastic episode of @hcrichardson.bsky.social I have also felt that the 1920's would define what MAGA feels the limits are, not Nixon The 1920's were defined by Woodrow Wilson, and it gets so unbelievably bad that the Nazis cite them as inspiration www.youtube.com/watch?v=UuKU...

it took a few days to get Sora to generate a video of a cat skating and landing a flip trick this experience caused me to think "ok... this videogen stuff isnt ready yet" and i havent really tried since generating video is very expensive, so ai-side mistakes are painful bsky.app/profile/jmsd...

i found it difficult to get sora to produce videos of skaters landing their tricks instead of falling never thought to try swapping the deck for a dach

one of the worst parts about AI coding is that so few READMEs have screenshots anymore I wont try anyone's app without being able to see what it looks like first i LOVE great design and miss being able to see the amazing things people are building by just visiting a URL

something very interesting about paradigm shifts is that it's usually not clear the new paradigm will replace the old one it looks like a toy for a while if the new paradigm proves itself, people start migrating, but sometimes new paradigms also don't go anywhere, even when they're better

i love my friends

I have my iOS UI Components library in fantastic shape I realized yesterday I could build a parallel version for Android And then use it to build native apps for both platforms that look and feel the same Gonna see how far I can get

my doormat shows how much i love wu tang

if my name were nick, id spell it knick

thinking out loud what if software can be built cheaply enough that entire video games could be used as instances of a meme example idea: imagine lots of people are able to build entire Strait of Hormuz simulators, with custom rules, and maybe like 2000 are made in a month. fps, xcom, rpg, etc

my friend at digital ocean is gonna slam tf out of memorymtrls i didnt build it with anything like this in mind but i cant wait to see how what happens

lol his shirt is so funny Stealing Until Nobody's Original youtu.be/erOMd8uESEk

oops i stayed up all night

deterministic guardrails hackernoon.com/system-defin...

I keep kicking around an idea that I feel must be in the future I dont claim it is good or bad From user POV, imagine I want to start each day with 3 songs that sound like NOFX (a fav band), but i want the lyrics to be yesterday's news I would get the news via 3 songs sound like NOFX in shower

shipped Borma v0.1.1 reviewed it with fable and fixed the nitpicks it found. the app is so simple that opus 4.8 did a fine job building it Borma: ideamtrls.com/borma/ On Testflight: testflight.apple.com/join/mMJgBZeC

🎸 Kids, by Descendents (guitar cover) playing this song on guitar is one of the things I do to keep my chops up the power chords switch so fast that it's like running down n backs youtu.be/mMBwGg62xB4

Why I think Anthropic's uneven safety policies with the release of Claude Fable 5 undermine the broader AI community's cohesion and accelerate us to more uncertainty and risk in AI's near-term evolution. www.interconnects.ai/p/claude-fab...

I'm not using too many tokens with Fable for my projects Given how expensive usage is for Fable, Im becoming increasingly fond of having memory be a simple set of docs that describe only *what is* for the repo, and throws away historical data This claude plugin is my first experiment

A+ take

It is truly wild to consider just how bad US infosec has been for so long. Every major company gets hacked. All the telecoms have been fully hacked. I became a jaded info hermit that resists ever giving info to anyone. And now AI might actually clean up the dumpster fire...

DARPA once created an electronic voting system based on zero-knowledge proofs voting *could* be fast and secure, if we wanted it enough www.darpa.mil/research/pro...

weird idea what if making software was so lightweight that you could make a zine for apps maybe a kind of app shell that is always your app for X, and you have a suite of these shells, like for email, calendar, etc and every three months you get an entirely new UI, eg new issue of the zine