Intro

I read a few things this week around AI and code and how things are going. I’ve been trying to think what this all might mean but I’ve not yet been able to put all my thoughts together in a coherent way.

Other than reading this stuff I’ve been generating more code to see where the tools take it. One I’ve been using is amp to help me with this newsletter by pulling together all the articles I’ve read and highlights I’ve made the last week. Not being a frontend developer means something like that would take me ages before, but now I can put it together in a couple of hours no problem.

Highlights of the Week

I Sent This Message to Everyone at Ardan This Morning

https://x.com/goinggodotnet/status/2012209293651501069/

I need everyone to start focusing on their engineering skills:

  • Being able to identify what needs to be built and why.
  • Breaking that down into chunks of work and that be verified and tested.
  • Knowing how to ask for the code you need written and when you need it.
  • Maintaining a mental model of what is being built and coded.
  • Code reviews by yourself and the AI coding agents.

Revisiting the iPod

https://paulstamatiou.com/revisiting-the-apple-ipod

It’s hard to imagine today what using an iPod in the early 2000s truly felt like. It was a dedicated device, singularly focused on helping you do one thing—play your favorite music—and do it well. It didn’t need to multitask or have any other complexity to function. It was simple, elegant and polished device. We don’t really have things like that today. Every mobile device is meant to be flexible, multitask and do just about anything for you, lending them to be more like a jack of all trades, master of none.

Year in review: Limitless, Sesame, Claude Code, and more

https://paulstamatiou.com/2025-year-in-review

The hard part now isn’t building. It’s restraint in this world where we have more capabilities than ever. I’ve noticed this industry is a bit too trigger-happy to jump into Claude Code and start generating, then immediately commit what was one-shotted. Maybe we should all be doing our own product plan mode outside of Claude Code’s plan mode. Taking a step back to ask what we’re actually building and why before we start. It’s too easy to skip the product thinking and strategy. That’s the necessary but slow and hard work of deciding what should exist, why, and who it’s for in the first place. What I’m referring to is just the normal product development process we’ve been doing for decades. Nothing new here. Just a reminder to not skip it when AI makes the building process closer and closer to a single button press.

New tools mean new ways of working and because the barrier to writing code or demos is gone, we need new ways of figuring out what it is we should build.

Agent Psychosis: Are We Going Insane?

https://lucumr.pocoo.org/2026/1/18/agent-psychosis/

It takes you a minute of prompting and waiting a few minutes for code to come out of it. But actually honestly reviewing a pull request takes many times longer than that. The asymmetry is completely brutal. Shooting up bad code is rude because you completely disregard the time of the maintainer. But everybody else is also creating AI-generated code, but maybe they passed the bar of it being good. So how can you possibly tell as a maintainer when it all looks the same? And as the person writing the issue or the PR, you felt good about it. Yet what you get back is frustration and rejection.

This is about open source but I feel is becoming more and more of a thing in companies too. Until now it has been about getting people to use the tools but as more and more do start using them we’re just all moving towards a new reality of cheap code generation, but the costs of everything else have remained the same. It is so easy to spin up a new app or service or bolt on a new feature, but reviewing, testing, maintaining and everything else with that code still has huge costs. Just because generating code is 10x cheaper now does not mean the rest of the system is magically gone away. We still need all those and they’re more important than ever now arguably.

Bring Back Ops Pride

https://charitydotwtf.substack.com/p/bring-back-ops-pride Another great article from Charity, lots I could quote from it so I’ll be selective but I do recommend reading the whole thing.

The cost and pain of developing software is approximately zero compared to the operational cost of maintaining it over time.

Reinforcing my earlier points of AI making code generation nearly free is not really a new thing. Writing code was never the issue. Everything else is where the majority of time is spent. This is doubly true in organisations where “ops” has become a dirty word that nobody wants to touch

The week ahead

The recurring theme of all this is that we need better tools as engineers for the full system, not just the code generation part. If all you want is POC or demo code then the current tools are great. Not so much if you’ve to maintain the software built by AI. Not yet anyway. We need the tools to catch up.