Programming

yeah, that's... one of the points in the article

We could see how the copilot PRs went:

Or to copy something and modify it 30 times.

This seems like a very bad idea. I think we just need more lisp and less AI.

Average drunk human coding skils

The average coder is a junior, due to the explosive growth of the field (similar as in some fast-growing nations the average age is very young). Thus what is average is far below what good code is.

On top of that, good code cannot be automatically identified by algorithms. Some very good codebases might look like bad at a superficial level. For example the code base of LMDB is very diffetent from what common style guidelines suggest, but it is actually a masterpiece which is widely used. And vice versa, it is not difficult to make crappy code look pretty.

what are the legal implications?

It would be so fucking nice if we could use AI to bypass copyright claims.

The early stages of a project is exactly where you should really think hard and long about what exactly you do want to achieve, what qualities you want the software to have, what are the detailed requirements, how you test them, and how the UI should look like. And from that, you derive the architecture.

AI is fucking useless at all of that.

In all complex planned activities, laying the right groundwork and foundations is essential for success. Software engineering is no different. You won't order a bricklayer apprentice to draw the plan for a new house.

And if your difficulty is in lacking detailed knowledge of a programming language, it might be - depending on the case ! - the best approach to write a first prototype in a language you know well, so that your head is free to think about the concerns listed in paragraph 1.

Counter-argument: If AI code was good, the owners would create official accounts to create contributions to open source, because they would be openly demonstrating how well it does. Instead all we have is Microsoft employees being forced to use and fight with Copilot on GitHub, publicly demonstrating how terrible AI is at writing code unsupervised.

Baldur Bjarnason (who hates AI slop) has posited precisely this:

My current theory is that the main difference between open source and closed source when it comes to the adoption of “AI” tools is that open source projects generally have to ship working code, whereas closed source only needs to ship code that runs.

Maybe closed source organizations are more willing to accept slop code that is bad but can barely work versus open source which won’t?

Because most software is internal to the organisation (therefore closed by definition) and never gets compared or used outside that organisation: Yes, I think that when that software barely works, it is taken as good enough and there's no incentive to put more effort to improve it.

My past year (and more) of programming business-internal applications have been characterised by upper management imperatives to “use Generative AI, and we expect that to make you nerd faster” without any effort spent to figure out whether there is any net improvement in the result.

Certainly there's no effort spent to determine whether it's a net drain on our time and on the quality of the result. Which everyone on our teams can see is the case. But we are pressured to continue using it anyway.

I'd argue the two aren't as different as you make them out to be. Both types of projects want a functional codebase, both have limited developer resources (communities need volunteers, business have a budget limit), and both can benefit greatly from the development process being sped up. Many development practices that are industry standard today started in the open source world (style guides and version control strategy to name two heavy hitters) and there's been some bleed through from the other direction as well (tool juggernauts like Atlassian having new open source alternatives made directly in response)

No project is immune to bad code, there's even a lot of bad code out there that was believed to be good at the time, it mostly worked, in retrospect we learn how bad it is, but no one wanted to fix it.

The end goals and proposes are for sure different between community passion projects and corporate financial driven projects. But the way you get there is more or less the same, and that's the crux of the articles argument: Historically open source and closed source have done the same thing, so why is this one tool usage so wildly different?

When did you last time decide to buy a car that barely drives?

And another thing, there are some tech companies that operate very short-term, like typical social media start-ups of which about 95% go bust within two years. But a lot of computing is very long term with code bases that are developed over many years.

The world only needs so many shopping list apps - and there exist enough of them that writing one is not profitable.

most software isn't public-facing at all (neither open source nor closed source), it's business-internal software (which runs a specific business and implements its business logic), so most of the people who are talking about coding with AI are also talking mainly about this kind of business-internal software.

There are commercial open source stuff too