Discussion
Hearing more and more stories of "nobody knows how to work on our codebase anymore after 1.5 years of agents, and management is asking for us to get more involved in the code after initially telling us to go agent native due to quality issues, but nobody really knows how to anymore"
Responses:
- Shock!
- Surprise!
- Wasn't this the goal?
@cwebber I keep thinking of that scene in Backdraft when De Niro is dressing down a fireman for doing something dangerously stupid:
"Was that before, or after, you noticed you were STANDING IN A POOL OF GASOLINE?!?"
That's the speech I would love to give to every AI forward leaning hotshot MBA cum CTO when they discover what they have done to their codebase.
My code cleanup rate starts at $175/hour. I'll be here all week. Tip your wait staff. ;)
Think of the consultancies like McKinsey & Accenture that will get hugely rewarded for recommending AI implementations, as a cost cutting measure... mass layoffs ensue...
.... in a few years those same shops will be re-hired because the ROI never materialized & the codebase is a nightmare no one internal will touch.... massively expensive re-writes ensue.
https://www.channeldive.com/news/accenture-ai-bubble-agentic-adoption-challenges/809668/
https://insights.som.yale.edu/insights/this-is-how-the-ai-bubble-bursts
If your source control person was at all on the job, there should be a tag like LAST_PRE_AI_VER.
@cwebber This is the kind of foresight that senior leadership gets paid for, I guess.
@cwebber perhaps they could simply ask the agents to tidy things up a little bit
@urlyman@mastodon.social @mcc@mastodon.social @aeva@mastodon.gamedev.place @cwebber@social.coop write good code, make no mistakes, make it easy to maintain, don't worsen climate change
@cwebber I remember calling this "technical debt as a service"
@cwebber Upcoming industry practice: AI weed-whacking.
@cwebber I love the smell of technical debt.
It smells like... employment for software archeologists!
@cwebber yea it's precisely the goal of LLM vendors. I'm *still* not clear why anyone without a large stake in a LLM company would have gone along with it though. Especially dependency averse megacorps.
My best guess is that my mental model of, "surely when trying it out you would carefully review the output for maintainability like Everyone. In. The. Industry. Has. Been. Saying. Is. Crucial. For. Decades." is not as universal as I thought it was.
@cwebber in my workplace we are still in the "embracing AI first" way of working stage. No amount of warnings seem to have any effect. I'm in malicious compliance -mode. Then I can tell I told you so. Though probably no one is interested in that either...
@cwebber Thou shalt beg for more, as you elevate me from .01, to .0001% wealthiest.
Yours,
#TechBros