In my line of work (IT Operations for a large international bank), we came up with a classification framework that tried to gauge information along two axes: Importance and Frequency of Use.
Importance is the more obvious one - there is critical info you absolutely need as a basis to understand the system, then there is also minor details. The difference is in an emergency, lack of the former will break your attempt to fix the system, lack of the latter will merely inconvenience you.
The Frequence of Use axis is the one that brought value to our operations when we introduced it. That's the question "How often will I need this info?". Regularly re-occuring, detailled procedures need to be accessible easily, so they go to a prominent Confluence page. But then there is information that you will hardly ever need, some arcanae about the system that only the old timers knew but that needed to be fixed in writing.
The most vexing bits of info are those that are highly important but are rarely ever needed. Do you plonk them into Confluence? Might not be the best idea because then you don't know how to quickly find it when need be. So we had an extra category for just that sort of thing which became our go-to info hub for all sorts of highly urgent but at first intractable issues with our system.
Part of the answer to this conundrum might be the "lightswitch effect": Lightswitches are ubiquitous and easy to use, so you stop thinking about them, take them completely for granted and miss the bigger picture that even 150 years ago, this would have been considered not a small technical miracle!
LLMs seem to fall squarely into that category - the threshold to using them is incredibly low because you can "just talk to them" and they give back meaningful answers loaded with the sum total of human knowledge from the interwebs.
If I may ask: In what areas have you found the most acceleration in productivity? For me as an enthusiastic amateur programmer (whose enthusiasm is, arguably, greater than my technical abilities) the only reliable and sustainable way to use AI in building software is to treat it like an exceptionally knowledgeable (though not perfect!), infinitely patient tutor that guides me through programming problems I run into.
The minute I relinquish control and let it just write code for me I am doomed because AI cannot fix its own mistakes in anything of medium complexity or above - and then I have to dig down and try to understand code I haven't written which is, in my opinion, worse than having had to write it myself in the first place.
I wouldn't consider myself an SWE, just an amateur programmer, so this is an honest question: In your opinion, is there such a case where the usage of AI in software engineering has either sped up the dev time or improved product reliability (or even both)?
I started using LLMs in 2022, quickly turned away because the things obivously made up stuff I am a certified expert it. Back to Stack Overflow. Came around to LLMs in late 2025 again as they rolled it out in my workplace and quickly succumbed to them becoming my all-knowing, (almost) flawless, infinitely patient teacher - that was the last time I opened Stack Overflow.
AI never tells me I'm dumb for asking a question, never passive-aggressively shoves genuine questions aside. And AI can just be as critical in dissecting my flawed approaches to problems as SO could be (which is good!) but without the toxicity.
Seems like the need for a support group depends on whether the recent advent of LLMs in your specific workplace has led you to become a Centaur ("I'm now a 10x SWE") or a Reverse Centaur ("this is rapidly sucking the joy this job once held for me"). So, yes, there is very definitely a need for this as there are definitely people seeing themselves cast as the latter.
Seems to me that both music and language fall into the same category of "patterns whose consumption will trigger a highly subjective experience (pleasurable or repulsive)". AIs as pattern matching machines can do this pretty well by now. Coding, however, is a separate category, one where cold, hard logic reigns supreme. Code is not consumed to illicit a subjective response but to work, fullfil its job - which might explain the gap between these two.
"...when a claim passes through five AI agents, how should trust propagate from one step to the next? I spent months trying to solve it."
Maybe I am naive here, but wouldn't a sane person try to turn a rational eye on the problem stated therein and - maybe -think "what if there is a way to avoid chaining of AI agents at all?".
"The company said the most likely explanation is that one of its transport partners sold the vehicle without removing K Group branding, despite contractual obligations requiring decals to be removed before resale."
I say this was done intentionally - the Iranian Revolutionary Guards surely can put that extra income from advertisement during a highly televised event to good use.
Hopefully it will fare better in its ecological and economical impact than the Assuan dam - which, I will never tire to say, has to put almost its entire electricity output to the production of fertilizers that no longer deposit naturally along the Nile's banks because of the dam(n).
This is a stunning idea and execution! I am glad I clicked on that link and spent a couple of minutes on it. Serious kudos, good Sir/Ma'am/[respectful appellation of your choosing].
Not sure if I fully qualify as a dev but I, for one, are rather confident in conversation. But with code, there is only two modes: Either I am the biggest idiot whoever disgraced Earth's surface or I am the absolute titanic Übergod for whom no problem is too hard to solve.
There is nothing in between which is pretty vexing.
Pretty sure their strategy is "hype it up, stuff it down EVERYONE's throat, once everyone is hooked and the tool is indispensable, make them pay for it. And rachet up the prices every quarter".