Hacker News Reader

Comment by BiteCode_dev | Hacker News Reader

There is another point to consider: more and more people are writing code without being experienced programmer, thanks to easier languages, better IDE, more streamlined lib API, and now, even ChatGPT.

A geographer scripting her GIS in Python or a quant doing some pandas analysis has a very limited programming experience, since her expertise is in her job.

And so we see more and more code produced by people that don't know what is slow and fast, what is safe or dangerous, what is maintainable or not. This code is popular, because it solves a real problem by people deeply familiar with the problem, way more than a programmer would be.

I have an example at $current_client, where they have some matlab code that handles tremendous amounts of money, with no comments, no doc, no tests, side effect everywhere and some interesting quadratic behaviors. They solved by loading balancing the thing on 6 servers with 17 processes, and manual testing.

View on HN

I do data engineering, and I'm working on carving out a niche for myself in optimizing code. Whereas most others want to solve sexy problems as soon as possible, it ends up biting us in the end because it doesn't scale down the road and/or costs tremendous amounts of money.

Our OKRs often include "reduce cost of X by 8%", which is nice but isn't achieved half the time. But if you have deep experience or are thinking outside the box, you can (as I have) reduce by 95%.

The code the unexperienced write is infinitely faster than the non-code that would be not-written by nobody would the unexperienced not write code.

Sometimes it's concretely a specific amount faster because you're replacing some other workflow with the software, but yes sometimes you're enabling something which would otherwise just not happen. Colleagues in our sister group, which doesn't have any actual software engineers, helped some academic librarians build a tool to automate something they knew they couldn't otherwise do at scale. They want to teach students how to use a library effectively, and while any of them could walk say, one student per hour through this, there are thousands of new students every year, it's not manageable. Midway through the presentation the specialist explaining it said "It's not a program" and I was glad to find I wasn't the only person who wanted to interrupt, because duh, of course it's a program. They've built a program. It's a bunch of instructions, for how the computer does stuff, it's a program.

Yes these people wouldn't have attempted Python, let alone Rust, but what they've built is a program, in a visual programming language. Is that intimidating and so they might struggle to maintain it? Yes, a little bit maybe, but pretending it isn't a program won't help. And that might mean sometimes you need a software engineer, the same way that if we want a bike shed (and sometimes we do, students like bicycles), sometimes we might need the staff architects down the corridor from me, don't just throw up a few brick walls and hope you got it right.

Well, it's a wild guess, of course. But does it sound even remotely possible, that thanks to that messy matlab code $current_client now has money to hire you to look into it?

Hence the "because it solves a real problem by people deeply familiar with the problem, way more than a programmer would be."

Oh, missed that on initial reading. Point taken.

Reply on news.ycombinator.com