"The problem with the Agile methodology" is that it doesn't exist. Somebody should have mentioned that before you "implemented" it.

"Well, Jason, if you spoke to some top-level engineers..." says the recruitment consultant.

I don't know who needs to hear this*, but it's not knowledge if nobody knows it.

* I do

@jasongorman +1 to that. Any mention of AI and I reject the job pitch. So that’s roughly 100% of LinkedIn adverts.

I’m patient. I can wait. But obviously if anyone has an opportunity for a senior ‘been there, done that, builds teams* that can deliver 20-30 drops/day without needing AI’ guy, then do get in touch. Contract preferred, but will consider full time for something interesting.

* the usual 7+/- 2. My record was an 8-person XP team delivering live 40 times/day regularly

On the question of using LLMs to build a "moat" for your business, ask your favourite frontier model 10x to pick a number between 1 and 10.

Now imagine *everybody's* LLM giving the same distribution of answers.

They're averaging machines, not moat-building machines.

I've been asked by several clients over the past year to help them find senior software engineering talent. You need to play down the "AI" rhetoric if you want to attract people who genuinely care about the work and who really know what they're doing.

Advertising that you intend to install them into some kind of "reverse centaur" role will pretty much guarantee they won't be applying.

When we get on to modular software design, I talk about "complexity by composition" - simple parts composed into complex and sophisticated wholes, and it's the interactions between those simple parts that produce the magic.

A symphony of FizzBuzzes.

As you can probably imagine, I'm not a fan of minimalism in music.

Having said that, this is one of my favourite pieces of music in the history of ever.

Complexity by composition.

https://www.youtube.com/watch?v=QA19NDIfXaQ

Large Language Models now face so many existential threats - technical, economic, legal, political, environmental, cultural - that if you're planning on models that are trained on accurate up-to-date information being around for the long term, you are probably a Large Language Model.

Just model collapse, as a single factor by itself, could do for them soon enough. LLMs trained on the agreeable, homogenised output of LLMs - the gen AI equivalent of Mad Cow Disease.

My brother recently switched me on to Unpeople. Debut album out now.

https://www.youtube.com/watch?v=HcJpjoivJ10

I know how excited software folks can get about anything with the word "dependable" in it

In the 15 years since I published my Dependable Dependencies Principle to help teams take a more informed approach to risk-based testing and refactoring, I've done a bunch more thunking about it. I finally got around to writing it down.

https://codemanship.wordpress.com/2026/10/13/revisiting-the-dependable-dependencies-principle/

"Open a PR on a popular repo today and observe what happens. Within seconds, a bot leaves an AI review summarizing your own change back to you. Another bot posts a preview deployment. Another posts a "walkthrough" with a sequence diagram nobody asked for. Sometimes a fourth shows up to respond to the third. By the time a human actually looks at it, the thread is a wall of generated text, and the real conversation (the human part) is buried somewhere in the middle.

That's the good case, where a human wrote the PR. The bad case is when nobody really did, like PRs that "fix" issues that don't exist, or PRs that rewrite half a file with confident, plausible, completely wrong code, or security reports that read like the real thing until you realize the vulnerability was hallucinated. And there are apparently accounts opening dozens of these a day across hundreds of repos, just farming green squares for a résumé.

I've started muting notifications from repos I used to love following since it's mostly just noise now, and that makes me sadder than I expected it to."
https://jross.me/open-source-as-we-know-it-is-dead/

For scale, that's about 4 seconds of my time.

"A feature that cost $2500 to deliver now just costs 25c".

How short was the meeting to discuss that feature?

You *did* discuss it, right?

RE: https://c.im/@cdarwin/117424109217801849

Will Elmo have to host his own "followers" now?

There are some of the usual suspects among the software factories "thought leaders".

The ones who told a mutual client they didn't need unit tests if they had Cucumber tests back when that was the hot new thing. Fast forward 2 years: "Our tests take 2 hours to run."

I look forward to cleaning up after them again this time.

Question for the folks who haven't surrendered all their thinking to Uber-Clippy: with static analysis, how might you identify Divergent Change?

Really enjoying the debut album from Unpeople, We Are Unpeople. Brother pointed me at it. Ordered. I think it's out today.

https://www.youtube.com/watch?v=3LCV-N-jdCM&list=PLESeT-jwV9b4&index=3

It's fascinating watching much of the media take at face value claims of future abundance for all made by the kind of people who famously don't let their employees take bathroom breaks.