
An escape route from YAML hell (News)
The Changelog: Software Development, Open Source
6min 45sec Sep 22, 2025
Adolfo Ochagavía believes we’re approaching the problem of configuration from a flawed starting point, Annie Mueller hits us with a wakeup call about how she reads beginner tutorials, Brian Kihoon Lee spends some time meditating on taste, Namanyay thinks vibe coding is coders braindead, and Can Elma speculates on why AI helps senior engineers more than juniors.
Changelog++ members save 1 minute on this episode because they made the ads disappear. Join today!
Sponsors:
- CodeRabbit – AI-native code reviews, built for the modern dev stack. — CodeRabbit is your always-on code reviewer—flagging hallucinations, surfacing smells, and enforcing standards, all without leaving your IDE or GitHub PRs. Trusted by top teams to ship better code, faster.
Start free at CodeRabbit.ai
Featuring:
Jerod Santo
What up, nerds? I'm Jerod and this is Changelog News for the week of Monday, September 22nd, 2025.
I would've published this week's episode earlier, but I got distracted trying to prove to Neal (from the awesome neal.fun wesbite) that I'm Not a Robot. I made it as far as Level 14, but it took me far too long to find Waldo. You?
Ok, let's get into this week's news.
An escape route from YAML hell
Adolfo Ochagavía believes we're approaching the problem of configuration from a flawed starting point:
> We are failing to see that configuration files are actually user interfaces, and that they should be treated as such.
>
> Once you start thinking of configuration files as user interfaces, it suddenly makes sense to demand an excellent user experience for working with them.
What might configuration software look like if our tools were rooted in the "configuration is UI" paradigm? Adolfo doesn't go so far as to present an "all-encompassing theory of what configuring software could look like in a perfect world", but he does shout out one particular project that he thinks is a big step in the right direction: KSON
How a non-dev reads beginner tutorials
Annie Mueller hits us with a wakeup call about how she, as a non-developer, reads the tutorial you, a developer, wrote for her, a beginner. Here's a sample:
> I first started doing Very Simple Thing2 with Snarfus, but the more I used it the more I saw the potential! Despite the jaggle of the chromus, it’s really multi-purpose. And that’s what led me to argyling the pintafore with the quagmire instead of the hoobastank! I know, crazy. But it was kind of working, and actually a lot of fun… Until I hit a big roadblock: the fisterfunk will NOT talk to the shamrock portal or even send beep-boops back to the Snarfus! Of course, you know what that means3 — Now the entire hoob-tunnel is clogged with gramelions. Unacceptable.
You already get the idea, but click through for even more snarfus, beep-boops, and ding-dongs.
Taste is humanity’s last edge over AI
Brian Kihoon Lee spends some time meditating on *taste*, along many dimensions. If it truly is "humanity's last edge over AI", then we should define and understand it for all that it is. Brian says taste is: art, parsimony, and the field behind the goalposts, into which the goalposts are constantly being shifted. Taste is cultural distillation, it's governance, and it's able to be refined. But how? He has some ideas about that as well.
It's now time for sponsored news!
Code reviews with context
The code review process has a history of being challenging for teams. It's where the work meets integration and, ultimately, solves customer problems. The process can go back and forth nitpicking syntax, style, or patterns, but without business requirements, deployment dependencies, or organizational knowledge... it’s only half the story.
That’s why CodeRabbit shipped MCP server integration. Until now, you could only bring context from Linear, Jira, or CircleCI. Now you can pull in any MCP server — Confluence docs, Datadog metrics, Slack discussions, even your internal tools. Instead of a shallow syntax check, CodeRabbit compiles all that context into reviews that understand what your code is actually trying to accomplish.
This makes CodeRabbit the first AI code review platform that orchestrates context across your entire development ecosystem.
And for our friends, who need integration at the coding agent level, CodeRabbit's CLI and VS Code integration does just that.
Teams can now enjoy code reviews that see the whole picture, not just the diff.
Learn all about it at coderabbit.ai or get your direct link to the blog post in this week's newsletter.
Vibe coding is creating braindead coders
Namanyay confesses to using Claude Code to write all of this code for him. And he thinks it's making him worse at the thing he's loved doing for twelve years.
> I can clearly see how AI coding is rewiring our brains – it makes developers crave instant gratification instead of deep understanding, and reduces us to gamblers who pull levers for the next hit of working code.
>
> If this is happening to me, someone who learned to code in the pre-AI era, what’s it doing to junior developers who’ve never known anything else?
Good question. It's probably too early to know, but I suspect the long-term impact of this change will be catastrophic.
Namanyay argues that when you vibe code, you get dopamine from the wrong source. He used to get two dopamine hits: one when he figured something out and a second when he got it to work. Now, the AI figures everything out and he's left with the shallow pleasure of things working. He knows the difference, because it's different than how he used to feel. But there's an entire generation of developers upcoming who will have never known anything else.
Why does AI mostly make seniors stronger?
This post by Can Elma pairs nicely with the previous one:
> The early narrative was that companies would need fewer seniors, and juniors together with AI could produce quality code. At least that’s what I kept seeing. But now, partly because AI hasn’t quite lived up to the hype, it looks like what companies actually need is not junior + AI, but senior + AI.
Can's explanation for why this is playing out is that AI coding tools are good at things juniors are generally good at (like cranking out code, trying different implementations, validating things, etc.) and bad at things seniors are generally good at (like code review, system architecture, spotting security issues, etc.). The end result, in his eyes:
> Instead of democratizing coding, AI right now has mostly concentrated power in the hands of experts. Expectations did not quite match reality. We will see what happens next. I am optimistic about AI’s future, but in the short run we should probably reset our expectations before they warp any further.
That's the news for now, but go and subscribe to the Changelog Newsletter for the full scoop of links worth clicking on. Such as:
- How to stop functional programming
- React won by default, killing frontend innovation
Get in on the newsletter at changelog.news.
We have some great episodes coming up this week! We visited our friends at Oxide Computer Company during their annual internal conference and we're shipping you a bunch of great conversations on both Wednesday *and* Friday.
Have a great week! Like, subscribe, and leave us a 5-star review if you dig the show, and I'll talk to you again real soon.
