kstrauser

From far down in the shady distance of the alley, a small voice shouts back: "Closed, duplicate!"

It's astonishing how vastly different our experiences were. I trawled the new Python questions frequently, looking for fun, novel things to answer. It was dismaying how often they'd been wrongly marked as dupes, often because the original answer was ancient and obsolete. It would often be egregiously bad, like:

Q: How do Python's dict preserve insertion order?

A: [duplicate of "Why does a Python dict's keys have a different ordering each run?"]

It was clear that the mod/curator/whatever spent no time attempting to understand what was actually being asked, and linked it to something that was completely incorrect as an answer to the real question.

I saw this constantly, so frequently that I'd be hard pressed to answer with specifics. If you asked me "when did you last eat a fruit bat", I can tell you the month and year because that stood out. If you asked me "when did you last eat spaghetti", I'd have a hard time saying.

Fair.

To be fair, there was a lot of X Y question asking there. "Would it be better to use MySQL or PostgreSQL to implement a clone of the Redis protocol? We're deploying an app that needs a 100% compatible Redis API. We own the Linux server it'll run on. Which should we use?" Uh, can you just use Redis? "Don't XY me!"

I'm not saying that's the problem you had, not at all. I do see why some of the answerers were quick to jump to it though.

How so? I stopped hanging around in Perl-land around 2001 and it seemed friendly enough way back then. What happened to it?

That infuriated me and I revoked those edits constantly, until I gave up and left the site altogether. I'd rather you steal my answer and put your name on it than put your own words under my name. Fix some dumb typo I made on mobile or something, sure, fine. Edit my answer to change the meaning or phrasing whatsoever? Oh hell no. Go make your own answer, and don't lie and say that I was the one speaking your words.

Right? It's not a very subtle spike, is it? Even then, if you erased the bump, the graph is pretty close to a straight downward-sloped line from 2016 to late 2022.

It's kind of sad: I bet there was some excitement around the SO office. Is this it? Are we on the way back? Have people remembered that we're here and now they're returning? Alas, no, they weren't, and they weren't. That must've been rough.

I did too! The gamification was fun, but more than that it was just really enjoyable to help other people. I like teaching, and SO was like the crack cocaine version of that.

Sure, but it rapidly fell before its pre-COVID levels and then kept dropping rapidly for the next year or so, and then AI really kicked in.

HN does a pretty good job of it. Turns out paid employees seem to be more incentivized to keep people coming back.

OK, let's get this out of the way:

1. SO question numbers dropped precipitously before AI came along.

2. AI clearly finished the job.

I took the graph from this very page and annotated it with the public launch of ChatGPT in November 2022.[0] Yes, the decline hastened after that, inarguably. But also, it was already nearly 50% down from its June 2020 local maxima. This is also inarguable.

SO had been flatlined since 2014 and declining since 2017. It dropped off a cliff when, 2.5 years after its 2020 bounce, ChatGPT came along with the death blow. Even then, it looks like 4 months of ChatGPT speed-ran about a year's worth of traffic decline, and then the slope went back to approximately the pre-AI angle.

I contend, and I'm backed by facts, that AI is not what started the avalanche. It killed any slim chance of a comeback, but that wasn't the beginning of the end. The end was already well begun by then.

[0] https://imgur.com/a/DN5N90L

You mean, the kind that made them more likely to sit around asking and answering questions because there wasn't anything else do to, and which caused an initial spike of traffic?

If there is, I’m not creative enough to imagine it. I don’t contend that SO could survive AI. You’ll never hear me arguing that. However, that graph shows that it was already dying before AI became widely available. AI 100% certainly kicked it while it was down. It just cannot have been the cause of the downturn because the downturn was in full swing before AI could’ve had any effect at all.

Traffic started dropping precipitously in summer ‘20. There were no widely known LLMs then, let alone ones used outside labs. ChatGPT went public in November 2022, well into the decline.

While AI was probably the last straw, it only hastened the collapse that was already underway.

Nah. It launched well, but sorely needed a BDFL. Almost without fail, services that delegate moderation duties to users are ruined. Inevitably, well-meaning people form councils and boards, and then those get taken over by people with nothing but time on their hands. It's the same thing that happens with school boards and HOAs: the last people you want running them are the ones that have the time and the drive to actually want to be in charge of them.

How is that different from clicking the Graph tab on that page? https://data.stackexchange.com/stackoverflow/query/1925371/n...

And here's why: https://honeypot.net/2026/10/09/what-really-killed-stack-ove...

LLMs are most commonly blamed, but the drop started way before those became popular. I'm confident that its horrid moderation drove people away long before AI had a chance to. Nothing's quite as demoralizing as:

Q: Why does adding two Python objects fails although their classes are compatible?

A: [closed as a duplicate of "Why do people say JavaScript is weakly typed?"]

I'm a reasonably bright person, with scads of StackOverflow fake-karma-Internet-points, but even I gave up trying to ask a question. There just aren't enough hours in the day to fight with hermits seeking the Ultimate Question and downvoting everything else.

This right here is why I hang around at HN.

Is JOE still your own daily driver, or do you use something else now?

Lawdy. Now we just need FoxPro Copilot to round it out.

Edit: No we don't oh please god no.

"Yo dawg, I hear you like security vulnerabilities..."

CF was an interesting path. I used the Python equivalent, Zope, extensively back in the day. Both could glue together some really powerful components and allow you to write whole web apps more or less declaratively.

That said, we collectively tried that experiment ("we" includes you, my JBoss compatriots) and realized this is not The Way Forward. That doesn't mean they're valueless, not at all! And it's the sort of experiment you have to try or else it's always going to be nagging at the back of your mind: "what if we tried this... and it was amazing?" But we did try it, and it was not amazing.

If you have to maintain CF or Zope or JBoss today, godspeed. Just know that life is easier elsewhere. It doesn't have to be this way. It's not this way for everyone.

…Server 2027.

Are you coding in Word or something? 36 point Times New Roman?

That kinda breaks my heart. I loved my kids curled up in my lap while I read them their favorite stories. I wish every parent and child had that shared experience, and I'm truly sorry you didn't.

If it comes to pass that you have little ones of your own, please don't let that be something your kids only see in the movies. Make that your own new family tradition.

I respect anyone who makes the tough call not to launch. There’s pressure to say it’ll be fine and we’ll just do it anyway, but someone sticks to their guns and says no, it’s not fine, and we can’t ship this. Doing the right thing in spite of the pressure deserves credit.

Well done, Fedora team.

Legit answer, to be sure! But if already on macOS, how does it compare? “Better and cross platform!” is the best possibility. “Nearly as good, but free!” is also solid. “Eh, it’s fine if you don’t have access to Soulver”, wouldn’t be a ringing endorsement.

For people who’ve used both: how’s it compare to Soulver?

Not that I recall and I’m feeling too lazy to look at the moment. It’s really the car’s fault, though. It’s sending the “start playing MP3s” command, which the device is obeying.

Hmmm, I’d have to see it. It’s basically the car saying “ooh you’re an MP3 player, play some music!”, and the phone is like, “oh, okay, you’re relaying your owner’s request so I’ll take your word for it”.

That's a "feature" of the vehicle's infotainment system. Some of the vehicles I've driven run Music the moment I connect my phone. Others are more respectful and wait for me to tell them to play something.

The aftermarket CarPlay adapter I installed in our Toyota is in the latter group. When I climb into the car, the CarPlay UI pops up on the vehicle's screen, ready for me to ask it to do something. Until I do, it does nothing other than show the homescreen with icons on it. A Ford I rented would fire up Music the second I plugged my phone in.

I’ve compared to building an asthma clinic on the top of Mt. Everest. If I could do that, I wouldn’t need to.