83 points?? If you gave me 100 guesses to figure out who scored 83 points I would never guess Bam Adebayo.
@mergesort.me
βApps geniusβ - Colleen. I used to work on making Twitter a bit healthier, then that all went to hell. Now I make puns, @plinky.app, and teach @ build.ms. Born and raised New Yorker, trying to do a little good and be the friend you made along the way. ππ±βΎοΈ
83 points?? If you gave me 100 guesses to figure out who scored 83 points I would never guess Bam Adebayo.
Last week ago I saw someone use e14n instead of enshittification and thought it was the funniest thing ever, a way to somehow make the term even shittier.
Yessssss. Say it louder for people in the back @simonwillison.net!
The #1 rule for being a Jets fan is that you will always do something nobody believes you should do.
Super cool! Thanks for letting me know. π
Fucking copyright. But I think my favorite line from the Wikipedia article has to be: The inclusion of Berry's "Johnny B. Goode" was controversial, with Lomax claiming that rock music was "adolescent", to which Sagan replied, "There are a lot of adolescents on the planet."
I started listening to this Beatles playlist two days ago and here came the sun, coincidence??
Every speaker gets a pair right? Just curious for no specific reason.
Every man has this thought and every man besides Bad Bunny is wrong.
Genuinely curious, can you tell me more about the project and how large and how custom things are? Plinky in particular is set up for agentic success (well-modularized with good APIs and Iβve been told by contractors βone of the cleanest codebases Iβve worked withβ), but there are still hitches.
Itβs still a super worthwhile trade off and I get a lot more work done this way. But this is where you really notice the difference between a high resource language (re: training data) like Typescript and a low resource language like Swift β in particular with newer frameworks like SwiftUI.
1/ I see lots of vibecoded SwiftUI apps out there and definitely think itβs possible to build one autonomously β Iβve made a few myself and love it. But thatβs why I think the reasonably complex part matters. I work in a ~100K line codebase and it struggles far more without extra tools and support.
I'd recommend trying to build something in a reasonably complex Swift codebase to test this claim. You'll find that without additional infrastructure around documentation, build systems, etc, Claude and Codex can struggle. I've built up this harness for my harness, but it's not great out the box.
Cats are funny lil buddies is the answer to all of βem.
Haha, I was just making a joke. But no real story, our cat is just a very adventurous little scamp! Hope the podcast is a good use of your time, and see you in Chicago. π
Meow?
A black cat sitting, masked under a black umbrella
Happy caturday, here's an umbrella and nothing more.
He's also a tremendous writer, truly a wizard with words.
To be fair that's not what he said and even X of all places has community notes that explain that.
I have a simple rule: I see a cat and I have to give them a like.
I ran a workshop today with a company that makes the app I've probably spend the most time in for the last 13 years β probably ~1,600 hours in that time β and I can't tell you how unbelievably rewarding it felt helping them with their AI knowledge and strategy so they can make the app better. π₯°
"I sure could use some more paperclips" - Me, if I was an agentic system
Love Is A Game, by Adele. (It was the last song played for Colleen and I at our wedding.)
Gotcha! Yeah, I use AI all the time to question my own assumptions β and then question those assumptions in turn because the process of thinking is what this is all about anyhow. π
Hmm, I can agree with some of this as not fully articulated in three tweet-sized posts, but I think itβs over-indexing on the idea of this being corporate vs. governance when Iβm trying to say that regardless of the context the questions ultimately boil down to civil liberty vs. collective safety.
Now Iβm curious, what are the edges?
There is no right answer, because no matter how the question is framed politically, the real question is: how much societal harm are you ok accepting to protect your civil liberties? Your preferred answer may be extreme or somewhere in the middle, but every solution will have meaningful tradeoffs.
The AI labs should build this feature. Oh, ok, I now have to trust OpenAI, great. No no, we'll have the government regulate it. Well I don't like this term's president and congress so that sounds bad to me. I can't trust anyone to make decisions besides me so we should have no guardrails. And so on.
Most societal matters of our day like age verification, AI content certification, or mandated encryption backdoors are primarily a question of where you want the locus of control in your life to live.
An app-walled article on Substack
Substack is now an email newsletter service where you can only read articles if you download their app.