AI Hot Takes From A Platform Engineer / SRE
19 points by yelianung
19 points by yelianung
Touches on "AI" with the same sentiment that Terence Eden's "I'm OK being left behind, thanks!" has.
https://shkspr.mobi/blog/2026/03/im-ok-being-left-behind-thanks/
Takes in this category seem rare, although they also seem completely logical. Nobody bothers to refute, they're just ignored. They sink without a trace. If these takes were wrong, people would cite facts. If they were logically flawed that would be pointed out instantly.
Instead, people talk around the central issue, which is that "AI" still kind of sucks, despite the immense amounts of money spent, and the management pressure to use it. Bizarre.
That post is showing a personal opinion. Even if I disagree with it in my own approach, why do we need to try to prove it wrong? We can let others have opposite takes without going all "destroyed with facts and logic!" in the comments.
If you want a comment overall:
For every HTML 2.0 you might have tried, you were just as likely to have got stuck in the dead-end of Flash.
People didn't get stuck with Flash. They learned things that generalise beyond it, created art communities, created things that we still remember and reference today. My friend got hired by another person from the flash community and is still working there. Flash dying didn't undo lots of things that Flash enabled. This can be generalised to many things.
I guess a part of it is the fatigue of dealing with the never-ending deluge of "My take on AI" posts. I think they're important, but there's just so much.
FWIW, the blog post you referred to did attract some (fair) criticism on Lobsters previously.
If AI coding companies truly think that their tools are the absolute shizzle, why is their commandline tool written in Javascript?
honestly, one of my longer term ambitions, now that I have seen LLMs become an actual usable coding tool, is to join in with whatever lot of open source enthusiasts starts using them to make software less resource hungry. I would love to see inefficiencies stripped away by new tools that can be a lot more thorough at identifying and fixing them than humans can.
Should we start a club? I've already done this for some tiny utilities. And every day I continue to use Slack, I'm tempted to vibe something native to the level of Ripcord, (it was taken from us too early) but open source. Ripcord, Shrugs and Volt seem to have stopped due to API changes breaking them, but with LLMs it's trivial - "here's a HAR file, fix the inputs". Ideally I want everything using Electron to DIAF and replace them with native apps that are fast. (Why does switching a channel in slack take 5+ seconds to fully load, ever?
Ah, also an AWS control panel which doesn't ****** **** ***** and take ages to load.
That and I'm already reverse engineering and opening some proprietary formats, because you can apparently do it automatically these days.
These were well projects where previously you could dedicate half your free time for years to solving one of them if they really bothered you a lot, but now you could realistically smash one out in a dedicated weekend.
Ripcord
Damn, never heard of this until today. I don't see much in the way of recent releases but I also don't see an end-of-life post. Is it really dead?
God I hate the Slack client...