A few days ago, I had dinner with a backend guy, and he complained that AI now writes code like it's cheating. Interfaces that used to take half a day to debug are now generated in seconds. But interestingly, he said he's actually spending more time on XML formatting lately. I laughed when I heard that. Isn't that an old relic from ten years ago? How did it make a comeback?
Then he showed me a project file, and I shut up. It was a set of microservice configurations. The logic in the AI-generated code was fine, but the XML configuration inside looked like a keyboard that had been stepped on by a cat—indentation all over the place, tags nested beyond recognition, comments and code mixed together. He said this was still on the better side. Sometimes AI loses namespaces or mixes up quotes in attribute values, which directly prevents the service from starting.
At that point, XML formatting tools become a lifesaver. Don't be fooled by how simple the feature looks; what it does is actually quite critical. First, it untangles that mess so the tag hierarchy is clear at a glance, easy for the human eye to scan, and quick to locate when problems arise. Second, it unifies style. AI-generated output sometimes uses two-space indentation, sometimes four, and attribute ordering is random. Once the formatting tool runs, the whole project immediately becomes orderly.
More importantly, XML is not as forgiving as JSON. If JSON is malformed, at worst parsing fails. If XML has unclosed tags or missing quotes around attributes, it really affects the whole system. AI writes code fast, yes, but it sometimes gets confused with strongly structured things like XML, especially when DTD or Schema validation is involved. Running it through a formatting tool at that point is like doing a static check, keeping many low-level errors out the door in advance.
There's another interesting scenario: merging configuration files. Many projects now use AI to assist in generating K8s configurations, Spring Bean definitions, or Android layout files. When XML fragments from different sources are pieced together, indentation and namespaces often clash. Formatting tools can realign these fragments so the whole file looks like it was written by one person.
So you see, the more powerful AI code generation becomes, the more XML it generates, and the more those XML files need to be standardized. Formatting tools have instead become infrastructure in the AI era, just like how the sharper the kitchen knife, the more important the cutting board. It doesn't produce code, but it makes code readable, editable, and maintainable. That's probably why a small, seemingly unremarkable tool has instead gained a firm foothold in the AI wave.