← Feed

Posts

• Post #1958309

  https://youtube.com/live/E95agtPgaa0 Who I really wish were there isn’t the compiler or Foundation engineers. It’s the app teams at Apple. Photos, Mail, Calendar, Weather and others. I’m sure a lot of those teams are dealing with the exact same challenges we are, and I’d love to hear what patterns and approaches they ended up using. It’s a shame we rarely get to hear from them, even though they’re probably the closest to our day-to-day pain.

• Post #1741798

  My girlfriend is shadowing the on-call engineer this week and her company is having a panic attack right now at 11pm because AI is down (apparently, there’s a mysterious billing issue). The CEO was on the call. Is this really the end of humanity? Have we outsourced our brains entirely? 😐 #ai

• Post #1741797

  - “Hey Claude, write me an app that does X.” - “No, redesign this part.” - “No, that’s wrong, change it.” - “Great, now you introduced a subtle bug. Do Y instead.” - “This is a race condition. Move this there and abstract it properly.” - “Write tests for Z. …Shit, I knew it. Let’s fix the logic.” “OMG I HAVEN’T WRITTEN A SINGLE LINE OF CODE SINCE DECEMBER.” “80% OF OUR CODE IS AI-GENERATED.” #ai

• Post #1741796

  It’s insane how pleasant using UIKit with an agent is when you have a specific UI/UX in mind. It takes a single detailed prompt to achieve EXACTLY the look and feel I want. I feel so dumb for all the time I wasted fighting with SwiftUI. 🙃

• Post #1421581

  “… we can entirely eliminate suspension points and make everything synchronous when dealing with an actor …” Only a MONSTER can leave me waiting for another week with this cliffhanger!!! 🥲 On a serious note, I’ve been really enjoying this series from @pointfreeco and cannot wait for the rest of videos! #SwiftConcurrency

• Post #893963

  I used to see previous LLM models get stuck when tackling hard problems, endlessly going in circles. I always thought they just needed to learn when to stop trying if they couldn’t find a solution. But that’s exactly what Opus 4.5/6 seem to do for me now, they quickly declare bankruptcy or try to convince me that the change is “too complicated” and shouldn’t be over-engineered (usually after reverting all my code). Turns out both behaviors are equally infuriating. #llm

• Post #893962

  @bootlessbuck It’s important to differentiate between the “figuring out what we’re building” phase and the final implementation phase. I’m talking about the former — where code quality doesn’t really matter because the POC will either be thrown away or heavily refactored. If you already know exactly what you want, then yes, giving the agent clear guidelines makes sense.

• Post #893961

  A funny possible future I’ve been thinking about is one where super complex, strict, and opinionated programming languages and architectures actually win. As annoying and difficult as they can be for humans to understand, they might be perfect for LLMs. We complain about the ridiculous complexity of Swift concurrency, but imagine a couple of years from now if it truly prevents race conditions at build time. That would be incredible for agents. TCA could be the next big thing too. #ai #swift

• Post #893960

  I don’t know which makes me want to cry more— that the US somehow thinks it owns the world, deciding which regimes live or die and playing god, even as it collapses from within, or that the governments of my own countries are helping the US fuck us over. Anyway, FUCK ISRAEL.

• Post #893959

  It’s really embarrassing that someone had to create a Chrome plugin to fix Claude’s bidirectional text handling. It’s 2026, and you still need to beg a 4 hundred billion company to understand that RTL languages exist in this world. Cannot Claude just vibe code that in Claude? https://chromewebstore.google.com/detail/claude-ai-rtl-transformer/pcnpnpaipomdildpaehlnmlbiiaagdid @claudeai #claude #rtl

• Post #893958

  https://parkerortolani.blog/2026/03/21/were-the-last-people-in.html I really miss Steve. I’d love to know what he would’ve thought about LLMs, and what he would’ve actually done with them. The world needs real visionaries. I’m tired of every mediocre product being labeled “revolutionary.”

• Post #450934

  The “writing specs” culture is becoming ridiculous. It was necessary when the cost of building was high. It feels productive when you do it. But if you think about it, you can create a POC with an agent much faster than you can draft specs. And that POC will answer most questions about how the experience should actually work, far better than trying to predict every detail upfront in a document. Product owners need to be closer to the creation. We need to change our mindset. #softwaredevelopment

• Post #310156

  RE: https://mastodon.social/@danielsaidi/116016801378997031 I used to really respect data-driven culture, and I still wonder whether there’s a healthy middle ground. But after working under it for so long, I’ve seen how it can slowly erode judgment, which is dangerous. Common sense starts to disappear, and product managers become slaves to the numbers, even when the right choice for users feels so obvious. It made me continuously outraged.