3 comments

[ 129 ms ] story [ 17.8 ms ] thread
So, the problem here is that we can’t trust the agents to get the code right. Right?

So, if we are going to the pull board method instead of pull requests, that means that all code has to have tests that it can pass or fail. Right?

And tests are code, too. Right?

So, if we can’t trust that they can get the code right, what makes you think that they can get the tests right? What makes you think that they can get the changes to the code right, so that they can get it to pass the tests? What makes you think that they won’t be changing the tests to match the code?

And what makes you think that they can test the code sufficiently to prove that it actually meets the bar being set?

What makes you think that they won’t be generating a bunch of random slop and then claiming that it passes the tests and meets the bar, and should automatically be checked in without any human review?

It seems to me like you’ve just built a random AI code slop factory, with an unlimited tokenmaxxing budget. Please show me where I am wrong, because I really don’t want this to be the direction that development is going.

Another HN article has highlighted the issue that each line of code is not actually an asset. It’s a liability. It’s just another place where your program can go wrong. I will quote one critical paragraph here:

> The illusion of productivity > > Watching code pile up is addictive because it feels like accumulating an asset. > However, code is a liability. The asset is the problem it solves for your customers. > The complexity of your codebase, and by extension the likelihood of bugs, grows > much faster than the code itself. That growth is mostly inevitable because > successful software tends to add features. Even so, the goal should be to keep > the complexity of the code roughly in line with the value it produces. If not, you > might find that any change starts taking longer even with coding agents. You > might also find yourself in an incident at 3 AM trying to decide whether you can > trust an AI-created bug-fix, for AI-created code, in a system you lost any real > intuition for.

See https://userubberduck.com/blog/software-intuition/ for more information.