We rewrote JSONata with AI in a day, saved $500k/year • Hacker News Board

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

#241

Post by dpark » Fri, Mar 27, 2026, 4:39 PM UTC

hperrin wrote: post #234

DrBazza wrote: post #154

Earlier quoted context omitted.

Most of the other replies to this hit the nail on the head. A human writing some poor, but working code that is supposed to be a demo, goes to production 9 times out of 10. Then it becomes critical infrastructure. Then management cannot understand why something working needs a rewrite because there's no tangible numbers attached to it. The timeless classic developer problem. We were here ^^^^ up to 2024-2025. Now, wi…

I don’t know where you got “likely correct” from. Likely working? Sure. Likely correct? Absolutely not. LLMs will only ever be as good as an average programmer, and average programmers usually get stuff wrong.

> LLMs will only ever be as good as an average programmer

What do you base this claim on?

> average programmers usually get stuff wrong.

All programmers get stuff wrong.

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

#242

Post by vitalikpie » Fri, Mar 27, 2026, 4:39 PM UTC

Well I'm seriously jealous about these posts. I rewrote this and that. One 10x engineer + Claude did everything in an hour.

It feels like I'm getting gaslighted.

I use AI at work with C#/Python - it fine. It can write some glue code and sometimes even pretty well. But I have to hand-hold it a lot.

My own project in Swift. Boy, AI can't handle Apple quirks - multiple iterations, code does not compile or missing crucial pieces (there are navigation links but not navigation stack).

I'm trying to be not picky. I want AI to do my job. But it's so far away.

Am I alone and everyone rewriting Linux in Rust over a weekend?

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

#243

Post by delecti » Fri, Mar 27, 2026, 4:46 PM UTC

bawolff wrote: post #13

I'm just kind of confused what took them so long. So it was costing 300k a year, plus causing deployment headaches, etc. But its a realitively simple tool from the looks of it. It seems like their are many competitors, some already written in go. Its kind of weird why they waited so long to do this. Why even need AI? This looks like the sort of thing you could port by hand in less than a week (possibly even in a day)…

My takeaway is almost the opposite. A company that has scaled to the point that they need 200 replicas of JSONata costing 300k/yr must be spending so much on compute that the difference is absolutely peanuts.

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

#244

Post by jmaker » Fri, Mar 27, 2026, 5:07 PM UTC

g19fanatic wrote: post #214

shafyy wrote: post #206

Earlier quoted context omitted.

I feel like 4.6 is worse than 4.5 lol

i actually agree for opus... but sonnet 4.6 is like magic ime

Whatever I used Sonnet 4.6 for, including Claude Code and Claude Chat, it made so many mistakes and totally awkward assumptions that I can’t fathom what it’s supposed to be good at. The mistakes were so blatant. Plan mode, several passes, couple grand in API costs… just disappointing at every task in every session over the past few weeks. Opus 4.6 has been good, still quite a few unexpected, silly mistakes, a few subtle but critical mistakes, but produced workable increments and code reviews, vastly subpar to GPT-5.x in chat mode (with and without identical customization).

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

#245

Post by macNchz » Fri, Mar 27, 2026, 5:20 PM UTC

danso wrote: post #229

macNchz wrote: post #216

Earlier quoted context omitted.

Opus 4.5 to 4.6 was pretty incremental, I didn't see much of a difference. The big coding model moments in recent recollection, IMO, were something like: - Sonnet 3.5 update in October 2024: ability to generate actually-working code using context from a codebase became genuinely feasible. - Claude 4 release in May 2025: big tool calling improvements meant that agentic editors like Claude Code could operate on a notic…

Maybe they're like me, who didn't spend a lot of time investigating Claude until 4.6 launched and the hype was enough to be the tipping point to invest energy. I do know that I've been having good/great results with Opus 4.6 and the CLI, but after an hour or so, it'll suddenly forget that the codebase has tab-formatted files and burn up my quota trying to figure out how to read text files. And apparently this snafu h…

There's definitely a productivity curve element to getting it to behave effectively within a given codebase. Certainly in the codebases I work with most frequently I find Claude will forget certain key aspects (how to run the tests or something) after a while and need a reminder, otherwise it gets into a loop like that trying to figure out how to do it from first principles with slightly incorrect commands.

I think a lot of the noise about letting Claude run for very extended periods involves relatively greenfield projects where the AI is going to be using tools and patterns and choices that are heavily represented in training data (unless you tell it not to), which I think are more likely to result in a codebase that lends itself to ongoing AI work. People also just exaggerate and talk about the one time doing that actually worked vs the 37 times Claude required more handholding.

The bigger problem I see with the "leave it running for the weekend" type work is that, even if it doesn't get caught up on something trivial like tabs vs spaces (glad we're keeping that one alive in the AI era, lol), it will accumulate bad decisions about project structure/architecture/design that become really annoying to untie, and that amount to a flavor of technical debt that makes it harder for agents themselves to continue to make forward progress. Lots of insidious little things: creating giant files that eventually create context problems, duplicating important methods willy nilly and modifying them independently so their implementations drift apart, writing tests that are..."designed to pass" in a way that creates a false sense of confidence when they're passing, and "forest for the trees" kind of issues where the AI gets the logic right inside a crucial method so it looks good at a glance, but it misses some kind of bigger picture flaw in the way the rest of the code actually uses that method.

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

#246

Post by kjksf » Fri, Mar 27, 2026, 5:24 PM UTC

hperrin wrote: post #236

kjksf wrote: post #230

Earlier quoted context omitted.

Fixing bugs is the goldilocks zone for ai. Especially if you have a test that the agent can use to test their fix. AI is not a junior developer, as some analogise, but Rain Man. Ultra autistic entity that can chew through way more logical conditions that you. As long as you can describe the bug well ai will likely fix it. Logs help. Let me give you specific example. Here's a fix made by claude to my SumatraPDF: https…

That fix is just the same code from earlier in that function pasted in again after another asynchronous procedure. I feel like you’re probably just a worse engineer than you think you are if you needed Claude for this.

There's nothing I love more than unfounded arrogance.

How about you try to make a change to SumatraPDF code base.

Let's see how good of an engineer you are when you actually have to write a line of C++ code in complex codebase as opposed to commenting on a check in with an explanation of the issue and a fix.

Claude fixed this crash in a minute: https://gist.github.com/kjk/d22af052499f70a45708c311eef201ff

Why don't you tell me, smart man, what the fix it and how long it took you to figure out.

If you can do it in less than a day, then we can talk about how better of an engineer you are than me.

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

#247

Post by kjksf » Fri, Mar 27, 2026, 6:00 PM UTC

chrismorgan wrote: post #148

captn3m0 wrote: post #2

For context, JSONata's reference implementation is 5.5k lines of javascript.

And no, there are no dependencies in package.json either (other than devDependencies for tests). Which cumulatively means a competent developer could probably port it in less than one day. They almost certainly spent longer working out how to deploy and integrate the original JS and ironing out the problems, than it would have taken to port it in the first place. That’s sad . And then they definitely spent much longe…

> port it in less than one day

There's confidence and there's barking mad delusion.

Here's the reality.

I once ported 50k loc from Java to Go. Here are details: https://blog.kowalczyk.info/article/19f2fe97f06a47c3b1f118fd...

Java => Go is easier than JavaScript => Go because languages are more similar. That was a very line-by-line port.

Because I was paid by hour I took detailed notes.

I spent 601 hours to port it.

50k / 601 = 83 lines ported per hour, 665 per 8 hour day, but really 500 per 6 working hours a day. No one does sustained 8 hours of writing code daily.

I would consider that very fast and yet order of magnitude slower than your 5.5 k per day.

10x is not a mis-estimation, it's a full blown delusion.

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

#248

Post by hperrin » Fri, Mar 27, 2026, 8:17 PM UTC

I rereleased this as public domain, since it’s all AI generated, so it is public domain:

https://github.com/hperrin/gnata

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

#249

Post by blharr » Fri, Mar 27, 2026, 8:53 PM UTC

andersmurphy wrote: post #228

9wzYQbTYsAIc wrote: post #177

Earlier quoted context omitted.

If the only thing LLMs did was clear enterprise technical debt backlogs, the end consumer would still benefit from the technology.

That assumes they don't accelerate the accumulation of technical debt. For each item cleared how many new ones are added. LLMs accelerate your good engineers and your bad ones. So the slop likely increase faster than it can be cleared.

And it will affect good engineers and turn them into worse engineers too

AI benefits rely on these good engineers having 5, 10, 20 years of experience pre-AI designing (and fully, thoroughly understanding) these systems. What's going to happen to that engineering skill after 15 years of AI use?

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

#250

Post by tzury » Fri, Mar 27, 2026, 9:01 PM UTC

One day a kid came home breathing heavily, to his father’s surprise face he tells, daddy daddy, I saved a dollar fifty!

How did you do it? ask the father

Instead of taking the bus, I ran after it all the way home.

If you were smarter, you could have save us $22 by running after a taxi!

This old joke came to mind while reading this post.

A tech company spends hundreds of thousands of dollars per year, “for years”, on a piece of software that could have been replaced by a month of coding top? (prior to LLM and all), you sit and write and save the money.

If I was an investor in this company I would have hire a team to look through their entire stack. See, if this JSON thingy alone is half a million a year, their entire cloud is at least $35MM annually.

Perhaps this is not even a bad business idea. One can offer companies to provide drop-in replacement for their costly “micro services that no one dares to touch” and share the cost savings.