This week’s online debate was about whether you should still bother to read AI-generated code.

Software engineers have never lacked things to argue about. Tabs vs. spaces. The length of your lines. Folders by feature or file type. Trailing commas. You’d be embarrassed to know how much time and energy has been spent on these topics across organizations.

There are practical reasons to care about clean code. Your human coworkers have to understand it, and someone has to maintain it.

I remember my first professional agony over clean code. I was working on a custom keyboard for Nike, and had to build a custom emoji picker (long press on a pair of sneakers, get a popup with different colors of the same shoe).

This wasn’t exactly a trivial piece of UI, especially in 2016. That popup is easy in the middle of the keyboard, but there are quite a few layout considerations and edge cases as you move towards the edges of your screen. Getting this to work well took a lot of logic and if/else cases. Was it pretty? Probably not. Did that picker feel right under your thumb? That was the part I gave a shit about. 

It probably won’t shock you that, a decade later, I rarely read my agent’s output line by line. But I spend a lot of time using what it builds. Finding the edge cases. Asking why an animation still doesn’t feel right.

That’s where my attention has shifted. Not away from craftsmanship, but toward the part someone else actually experiences.

The person using the Nike keyboard didn’t care how many if/else statements it took. They cared that when they pressed and held a sneaker in the corner of the screen, the emoji picker opened exactly where they expected it to.

The code didn’t have to be perfect. Picking a pair of sneakers did.

To craftsmanship,

Jeremy

New models, new apps, new ways to spend $500: Dots, GPT-6.1 Sol, and another round of subscription announcements. OpenAI packed a ton of updates into DevDay. Dots bring always-on agents with their own cloud computer, and you can use your ChatGPT subscription in supported third-party apps through Sign In with ChatGPT.

Argon enters the chat: Google previewed Gemini 4 Argon for long-running coding and cybersecurity work. While the benchmark scores are positive, Bloomberg reports employee skepticism about the new model.

Your next video call could get weird: Tavus says 48% of participants in its own 54-person study thought its Griffin video partner was human after a one-minute call. We didn’t learn anything from the movie HER, so here comes HIM (Human Interaction Models).

Hop on the decision model train: Your agent has more ways to make up its mind. OpenAI’s Decisions API puts Luna behind predefined choices. Cloudflare’s Clef brings open-source multimodal models to Workers AI, Perplexity adds a hosted Decisions API, and Strands Decider 2B runs locally.

Now that’s what I call a loop: Have one model review another’s work. Uncle Bob shares how he uses Codex to review Grok’s code.

Pimp my Claude: Claude Mods let you change prompts, tool behavior, commands, and the interface. Yes, you can ask Claude to build them.

Stop grading output on vibes: A guide to having Claude build evaluations for your app, then improve it one change at a time.

New week, new agent fleet: Researchers have spotted a fleet of AI agents running on Tencent infrastructure and using a website-scanning service to query Alibaba’s Amap for directions.

Agent swarms could be AI’s next scaling frontier: What if the next leap in AI comes from models working together, not just getting bigger? Understanding AI explores how agent swarms could turn more computing power into better results by dividing work among dozens - or thousands - of collaborators.

Astra enters its Napoleon era: Carter Church used GPT-6 Astra to decipher a 217-year-old military letter to Napoleon’s general. His tally: six hours of model execution, plus a few evenings of his own. Transcription, a custom solver, historical cross-checks. A pretty good side quest.

You’re up next, accounting: Mercor published a new study, putting agents and 12 CPAs through four structured accounting tasks. Frontier models hit 100% of the grading criteria; the unaided human average was about 37%. But please, don’t fire your accountant over an Elon repost.

Everybody hates AI: The industry promised abundance. Derek Thompson argues that, so far, the buildout is making life more expensive. So while people use AI, they don’t trust it. The gap between the promise and the experience matters. While we may be on the verge of a productivity revolution, people will judge AI on whether it makes their lives better.