The plan to deal with the tech debt:
There’s a ticket in the backlog. We’ll get to it next sprint (aka never).
A recent job I had, Product kept saying this over 2ish years…
…at the end of the stint and with ~2 weeks until launch, they laid off all but one person on the team (some 12 engineers or so, myself included).
I often wonder how much shit that person is wading through leading up to launch and post-launch fighting fires.
If they are smart they would have bailed as soon as they fiund another job once everyone else else was let go.
But they probably kept the one person they knew was afraid of quitting.
Afraid of quitting not so much, but they are underpaid for their skills. At least for what they need the person was more skilled than me, but being paid ~$40k less. So…yeah…
Step 1: AI it all
Step 2: ?
Step 3: Profit!The AI will detech the debt.
It’s the only plan form all the teams that the company actually executes!
A study found AI is slowing experienced developers down: https://www.reuters.com/business/ai-slows-down-some-experienced-software-developers-study-finds-2025-07-10/
I’m awaiting the fireworks for when they use AI to patch the Cobol systems still used by the banks.
Just keep writing on the same line ? Fixes everything
Exactly, just run your magical new code in a minifier and problem solved! It’ll be just as speedy as the real deal!
Ok… just for you, I will extend the saying to be “every line and column is a liability”.
It sounds like you’re suggesting single-letter variable/function/class names everywhere
I don’t see why… once you “buy a column” (which you must weigh the trade-off towards readability), subsequent uses of that column on other lines are free (save the line itself, of course).
Every codepage is a liable as well.
Bill Atkinson and his -2,000 lines of code would like a word
Even then, so the theory goes… every line of code is a liability, it is only emergent properties of the system as a whole that makes it an asset. It takes but one line to destroy it’s value, and in general a 2kloc codebase is more valuable than a 4kloc codebase, if they do the same thing. QED? :)
Is it objectively more valuable?
I’d say if performance and functionality remain the same it improve and complexity is reduced then the 2kloc codebase is better
If it’s 2k lines of perl, it’s worse
Though in his case it sounds like he reduced complexity and improved performance, so definitely better.
I don’t know if I’ve even written -2000 lines at once, but I’ve definitely done -700 or so at least. Feels so good to delete a bunch of code without any loss in functionality.
