Over the last year or so, I’ve made the biggest-ever change to the way I write software. I now code with AI assistance all the time. Here’s why. Here’s how. kocienda.micro.blog/2025/06/25/c...

We’ve been making great progress at Infactory, the company I co-founded last year. We help developers build natural language APIs over structured data: connect data sources, build query programs, and deploy them as REST endpoints. infactory.ai/developers

How do you solve a design challenge full of unknown unknowns? Jump back to 2005 and try the Blob Keyboard, @kocienda.bsky.social's first working prototype for the iPhone keyboard, created during a UX crisis.

A robot garbage man who was never programmed to dream but does so anyway. #gpt4o

When I add big new features to code, I hack away wildly in my work branch. I'll often break other features. I don't care. I just want to see the new feature ASAP. Once I have that, I usually find it pretty easy to go back, fix any breakages, and neaten things up.

There are “Esperanto” ideas, proposals which sound great, but are complex and impractical, and so they never take off. There also are “tweet” ideas, trivial things which don’t seem to do much at all, but are simple and appealing to everyone, and so they spread everywhere.

I’ve been doing a bunch of coding over the past few days with claude-3.7-sonnet MAX in Agent mode in Cursor. It’s been amazing. It’s like having a junior dev who works fast, knows a lot for its experience level, gets a lot right (and a lot wrong too), but is always willing to go back and try again.

I’m proud to have started Infactory with Brooke, and I think we’re making great progress, but I know it’s rarely easy for her. Let her tell you. www.linkedin.com/posts/abhart...

I haven't written software for Apple platforms since 2020, but over the last few days, I've been working on an iOS demo app for a prospective customer. In an Xcode log message, I was unexpectedly greeted by an old friend: UIKeyboardLayoutStar. I wrote the first version of that class back in ~2008.

I have a “software mindset”. When I think about ideas or look at systems, I imagine the abstractions, data, and algorithms I would need to write a relevant program, either to examine a topic, simulate a real-world thing, or actually implement some working code.

no ulus for you

It’s 6:30 AM. In my favorite Hawaiian spot. Already finished a leisurely breakfast. Beautiful view of the surf. Laptop opened. A wonderful day ahead of programming ahead of me, since that’s what I like to do best.

I do a lot of python programming these days, and sadly, there’s no consistency anywhere. Let’s take just one example, is it: len(foo), foo.len, foo.length, foo.len(), or foo.length()? I never know.

I enjoyed every day I worked with Apple ObjC frameworks. Once you learned the naming conventions, you could guess names even in new frameworks. It made it easy to think in that code.

Delicious!

Merry Christmas to my internet friends. 🎄

There’s a fine line between software that’s supple and helpful without getting in your way and guessing wrong all the time, but great software comes along with you for the ride, and makes the ride a pleasure.

We can, and probably should, take it for granted that all decent software gets the algorithms right. Getting the heuristics right is what makes software excellent. I wrote this about heuristics in my book.

We applied Postel’s Law throughout the software for the first iPhone, and we described the concept in the ’949 patent under this title: Touch screen device, method, and graphical user interface for determining commands by applying heuristics. patents.google.com/patent/US747...

The best software gives what you want, not necessarily what you did. This is a corollary of the more general Robustness Principle, which I always refer to as Postel’s Law: “Be conservative in what you do, be liberal in what you accept from others.” en.wikipedia.org/wiki/Robustn...

proud to be the first follower of @jeangordonkocienda.bsky.social, the best person I know ❤️

Pt 2 of @kocienda.bsky.social's podcast with Jerry Cuomo is here! Hear all about Ken's journey after Apple, thoughts on AI, & what he's up to now at Infactory www.youtube.com/watch?v=t-s_...

Two rules of software: 1. Add a level of indirection if the code doesn't do what you want. 2. If that didn't work, goto 1.

I was on the Wild Ducks podcast with Jerry Cuomo talking about Apple and autocorrect. A fun conversation. Future episode on Infactory (my company) and AI will drop soon. open.spotify.com/episode/3IH8...

Waymo is still super cool every time. Feels like magic.

on this day in 1968! so good. en.wikipedia.org/wiki/The_Mot...

We’ve kept the Rancor out in the living room ever since. Around Christmas time, he gets a hat. Rancor Claus! He thinks you look delicious.

In 2009, I bought this Rancor toy. We had this gingerbread house building contest, and I decided to make a monster movie tableau. I thought it came out great! I made a 1950s style monster movie poster for it too! I didn’t win. 😅

At Apple, I wrote a little declarative language to describe different keyboard layouts. It handled geometry, keyplanes, key actions, graphics, colors, etc. It was cool. I called it KBStar. I wonder if they still use it?

Of course, LLMs make this possible in a way that I didn’t think I would see in my lifetime. And while talking to the computer in English in amazing, and is a big part of what we’re trying to accomplish at Infactory (the company I started earlier this year), that’s not the same as programming.

I’ve always loved writing little domain-specific computer languages. It’s fascinating to figure out the most economical way to say the thing you want the computer to do for you.

Over the past two years, and at an accelerating rate over the past few months, I write software with the help of AI. I come up with the ideas, write the specs, and nudge things in the right direction. The AI does the drudgery of putting the bits in the right order for me to review and integrate.

I like it the most when the new software teaches me new things. Doing a lot of work with Polars lately. Nice software. pola.rs

For me, the best programs to write are the ones where I know what I want to do in the big picture, but I’m not exactly sure how to accomplish the thing.

I’m happiest in my work when I’m writing new software every day, and I’ve been doing that recently, so 😃.

OK. No more programming for this evening. Honestly, I can stop anytime I want.

my fav ornament

I’m in NYC this week and yeah that totally was Alec Baldwin I passed on the sidewalk while I was walking to get my morning bagel and coffee. Small city. Small world.

I’m an extrovert with objects and an introvert with people.

I want computers to get a lot better which also means I want them to be different than they are today.

The people with the best ideas are usually the ones with the most ideas.

these days my morning commute music is silent thoughts mixed with the wail of a porsche flat six

Lately I’ve been cross-posting my tweets/toots/skeets on three services. I don’t imagine I’ll keep doing so for long. I just want a place to write little text things and have people see them if they follow me. No desire for the service to be part of the story. But here we are.

GPT-4 is great at writing AppleScript which is wonderful since it’s a tricky language to work with. The AI really helped me with this cool little demo I’ve been making over the past couple of days.

If someone claims to know a lot about design, what have they designed? If they want to tell you about how to make great products, what great products have they made?

I try to do great work, and for guidance and inspiration, I look to people who have done great work themselves. I read and listen to them, how they describe what they do, and how they achieve their results. I look to people who have proved their ideas and processes actually can succeed.

A few times, someone on the internet who wasn’t there claims to have knowledge and insights into the work I’ve done and can somehow correct the story on how things happened on the iPhone project because they understand it more deeply than I do and how I wrote about in my book.

I started on the iPhone project in July 2005. There were days when it was unclear whether it would work at all, yet there also was a belief in the incredible potential of the product if we could get some important basics right. We did and that belief turned out to be true.

Six years ago today was my last day at Apple but also the first day on the projects I’ve focused on since then. Excited about the future! 🚀

So few pre-web print magazines have survived through to today while plenty of newspapers did. Printed books made it through too, but bookstores mostly didn’t. Crazy.