We rewrote JSONata with AI in a day, saved $500k/year

> At Reco, we have a policy engine that evaluates JSONata expressions against every message in our data pipeline - billions of events, on thousands of distinct expressions.

The original architecture choice and price almost gave me a brain aneurysm, but the "build it with AI" solution is also under-considered.

This looks like a perfect candidate for existing, high quality, high performance, production grade solutions such quamina (independent successor to aws/event-ruler, and ancestor to quamina-rs).

There's going to be a lot of "we were doing something stupid and we solved it by doing something stupid with AI [LLM code]" in our near future. :-|

No offence, but inexperienced JS fanatics always do this because of some weird affectionado they have for the language itself. Otherwise, even a decently qualified CTO would have chosen to keep everything in Go from the beginning or might have not waited until they were bleeding $300k. JS is also the worst possible language choice for this problem. So, it definitely sounds a bunch of script kiddies with fancy titles bought with VC money rather than actual experience.

The headline seems to be flashy indeed, but ai didn't really solve this imo.

They just seemed to fix their technology choices and got the benefits.

There's existing golang versions of jsonata, so this could have been achieved with those libraries too in theory. There's nothing written about why the existing libraries aren't good enough and why a new one needed to be written. Usually you need to do some due diligence in this area, but no mentions of it in this post

In order to measure the real efficiency, gnata should've been benchmarked against the existing golang libraries. For all we know, the ai implementation is much slower.

The benchmarks in the blog are also weird. The measurement is done within the app, but you're meant to measure the calls within the library itself (e.g calling the js version in its isolated benchmark vs go version in its isolated benchmark). So you don't actually know what the actual performance of the ai written version is?

The only benefit, again, is that they fixed their existing bad technology choice, and based on what is observed, with a lesser bad technology choice. Then it's layered with clickbait marketing titles for others to read.

I'll probably need to expect more of these types of posts in the future.

But if the ai built solution is slightly less stupid, then it's still a win isnt it?

> then pointed AI at it and had it implement code until every test passed.

You used to have two problems. Now you have three.