Stack Overflow, the programming Q&A website founded in 2009, set out to be different. To "make the internet a better place". Faced with endless forums full of technical suggestions of dubious authority, and with the next best thing paywalled off, Joel Spolsky and Jeff Atwood wanted to create a community focused solely on building a repository of quality content. No chatter, no to-and-fro: just questions, and answers. For a while, it was wonderful, and spawned a number of similar sites on other, non-programming subjects, all under the "Stack Exchange" umbrella and managed by Stack Overflow, Inc.

Naturally, over time, things began to change. As any useful resource grows in popularity, the community was faced increasingly with the natural ramifications of a simple fact: those who know nothing ask more (not to be misquoted as "those who know less ask more", which isn't true). A few years after its launch, Stack Overflow was known the world over as a place where you could get your problem solved for free with a minimum of effort. And, though many domain "experts" remained, asking complex questions on difficult topics for the community to solve together, they increasingly found themselves grinding, simply asking the same basic, introductory questions time and time again. And why not? That's helpful, right! To this day some still don't mind doing that, so the site remains known for its low barrier to entry, and so the feedback loop continues.

However, this had a distinctly negative effect on the community. Those who still treated Stack Overflow as the pure resource as which it had originally been envisioned, were becoming increasingly disillusioned. This manifested as frustration in communications on the site, and an increasing perception of hostility to new users developed. Many to this day consider Stack Overflow to be a horrible place to be, and will effortlessly spam out tales of how victimised they felt after their first experiences there (while ironically omitting the fact that they'd, in the meantime, indeed had their problem solved for free by volunteers).

Compounding the issue was a similar change in the views of the team running Stack Overflow. The first signs were just evidence that the basic principles of the site were no longer taken for granted by the staff. This, while disappointing, was really just a natural effect of the company's growth and of the exponential growth in the community's userbase (again, mostly newcomers to the industry, with the slant on question "interestingness" that this would bring). A few attempts were made at diversifying the product; none of them were successful ("Stack Overflow TV" evaporated, an attempt at deploying volunteer, female-only mentors to a for-profit bootcamp may or may not have happened, and the absolute disaster that was "Documentation" immediately comes to mind).

Some of the most well-known experts decided that the site wasn't quite what they'd signed up for, and decided to reduce their level of contribution, or leave entirely. At the time, this was fairly big news. People generally wanted the company to focus on being good at one thing, not terrible at ten things.

Eventually, in response to the worsening reputation of the site, and feeling ever-increasing pressure to begin providing returns to the investors who had backed the original offering, Stack Overflow began reshuffling its management team, increasing its focus on revenue generation by increasing the focus on "Teams", a paid-for spin-off designed for use in-house by companies with a desire for private Q&A. This offering would ultimately be advertised on one of SO's own sites, in the guise of "sponsorship" (how can you sponsor yourself??), and around the same time drove a complete overhaul of Stack Overflow's front-page, which many pointed out was not just a case study in poor website construction but also borderline offensive to those who still saw themselves as willingly contributing to a completely different kind of platform.

Adverts were introduced to all users as an "experiment"; promises that these would never include animations (or, more broadly, potentially dangerous scripts) were swiftly disproven; after it transpired that this prohibition was enforced not on a fundamental technological level but by reactive human oversight, and the notion was eventually passively retracted (despite the impact on accessibility — this, it must be said, not for the first time).

"Welcoming", the decline of Meta, & your data

Several "incidents" then transpired over the years, many of which left the community feeling like the company was listening to them less and less.

A proactive emphasis on being "welcoming", a well-intentioned attempt to improve the site's reputation to passing visitors, apparently had the opposite effect on contributors, many of whom wondered aloud whether the site was still right for them after being falsely accused of various evil deeds.

New director of Q&A, Sara Chipps, later announced that the staff would be paying little to no attention to Meta henceforth, due to a feeling amongst the team that toxicity had reached unacceptable levels. Ironically, said "toxicity" generally took the form of high-volume contributors expressing concerns that the company was no longer listening to them; still, Sara seemed convinced that she needed to shelter her staff from the hordes of evil internet warriors, and that would be that.

What had originally been billed as a site that "is built and run by you" was seemingly no longer such a thing, though that wording remains on the site to this day, despite all that was still to come.

Then came two data breaches: first, in January 2019, SO inadvertently leaked over a hundred user email addresses by hitting the wrong button in their personal Gmail. Then, in order to implement a small prize draw in January, they handed several hundred email addresses to Amazon, in contrast to not only good practice, but also good ethics, privacy codes of conduct and (at least in Europe), quite possibly the law … solely to avoid a small amount of work in the Marketing team. These developments were fairly shocking; still, nothing really came of any of it (except for an angry denial that it had counted as a data breach in the first place), and people moved on, growing wary of all this, but otherwise mostly content to put the past behind them as long as such a thing never happened again.

Illegal relicensing

A few months later, at the beginning of September 2019, Stack Overflow suddenly announced that all the freely-provided content on the site would now be licensed under CC-BY-SA 4.0, a popular and reasonable Creative Commons licence. This seemed to be done by accident to begin with, then made official.

The problem is, and this should frankly be obvious to anyone with even a passing comprehension of what licences actually are, they do not have any legal right to do this. When volunteers provided content to the site, they did so under the terms of CC BY-SA 3.0, not CC BY-SA 4.0. No matter how many people think 4.0 is a superior licence, you cannot simply decide that somebody else's works are now covered by a different licence. This is confirmed by Creative Commons themselves, who go on to explain that this act effectively nullified Stack Overflow's rights to use the content at all. For a site whose very existence is predicated on the very existence of (and continued submission of) such content, that's surely troubling.

Despite this, whether Stack Overflow disagree with the premise of the userbase's complaints, or simply don't care (either of which is pretty terrible), when challenged on this immediately by a large number of users, Stack Overflow opted not to respond, for some three weeks. At all. Not a single word on the topic. Eventually a staff member did post some boilerplate, which did not address the issue at hand. To date, there has still been no further engagement from the company. And that's arguably even worse than the original mistake.

Jeff Atwood, long since retired from the organisation but still present once in a while, handily asked people not to "forget" that a relicensing from 2.5 to 3.0 had occurred several years ago, and suggesting that everybody should "get some context, folks!". As if the fact that an illegal act had taken place before should make it okay now; besides, users on that very page had indeed pointed this out some three weeks prior. To me, this was the first sign that the problems now showing up had perhaps been present behind the scenes more generally for some time.

Meanwhile, in a further departure from normal operation, staff members were deleting posts, un-"starring" chat messages to make them less prominent, and manually removing "bounty" texts that seemed critical of the company's handling of the situation, and there was perhaps even some evidence that Stack Overflow had asked the Internet Archive to remove old pages to further censor the older text (though this evidence remains inconclusive).

At the time, this was perhaps one of the biggest controversies the site had seen. The announcement was the lowest-ranked on "Meta" (a sort of "Q&A-ception" where, at least historically, people would discuss and form policy about SO itself) of all time, and the rebuttals were some of the highest ranked by contrast. It didn't seem like this sequence of controversies could be beaten in its infamy for years.

Instead, it would take just three weeks.

Pronouns, and the great moderator exodus

On 29th September, it transpired that a well-liked and well-respected moderator had been summarily dismissed by the company, in the hours before a religious holiday, without the usual moderator dismissal process being followed, because in a private moderator chatroom she had politely queried the sense involved in a specific clause of a yet-to-be-implemented change to the website's code of conduct, as it related to pronouns, of all things. She had tried to communicate with Stack Overflow various times, and had no response until discovering she had been removed from "office".

The storm that ensued was nothing short of monumental. Seventeen moderators resigned in protest, and Meta's downvote record was broken again. Meta had suddenly become, primarily, a discussion forum for the subject of community trust in the company's leadership. Some users took the opportunity to employ sarcasm; others were more direct.

Staff demoted posts that seemed critical of their actions, despite the earlier announcement that they would no longer manage Meta, instead leaving it to the elected volunteer moderators (perhaps because so few remained). I suddenly missed the days when the biggest power shift I'd encountered was a moderator single-handedly nuking questions because he personally didn't find them useful (which, at the time, was generally seen as a task to be undertaken by the community, by consensus, using the tools that are still in place today).

Finally, the company's CTO David Fullerton, just this evening, performed some much needed damage control, addressing the moderator firing. Although no other topics (such as the relicensing) were touched on, more generally the post showed belated awareness that it is not enough to shout and scream at the creators of your product: you have to engage with them in at least some constructive way. Hopefully, as a result, we find ourselves at the start of the road to repair.

So what now?

It is hard to guess how things will go from here. Will enough experts vacate the premises that only a husk remains, a skeleton crew repeatedly answering the same "debug my code" posts with no further ongoing value remaining? Or will it all just blow over? There is still the spectre of a potential lawsuit over the illegal relicensing fiasco, which to date goes entirely unaddressed by the firm. And, although unlikely, technically someone could fork the whole thing and try again.

In perhaps a cruel twist of fate, Spolsky just this year resigned as CEO, handing the reins over to one Prashanth Chandrasekar. It remains to be seen whether the new boss will take the firm in a different direction, or steer it back to its roots, or even double down on the direction that it's been going in over the last few years.

One might also wonder why any of this matters so much. It's just a website, right? Go outside, enjoy the sun. To some degree that is a reasonable argument. However, it's tone-deaf to the millions of hours invested by its most dedicated users, to the millions of industry professionals now absolutely reliant on SO as an information resource (be that right or wrong), and of course to the people whose paycheck is signed by Joel Spolsky, who remains at the company as Chairman. The one thing that I believe most people can agree with is that nobody wants the site to fail. But at least one person is going to have to ask, then answer, the very important question: how can it succeed?


Update for November: Lawsuits and Time Travel

Sadly, Fullerton's promises never came to much. Those who were filled with much-needed optimism by his post, and upvoted it into the hundreds as a result, changed their minds after a few weeks when no further attempts were made to rebuild trust, and when it became evident that Monica's case was still not being heard. As of this writing, she is crowdfunding a defamation lawsuit against the firm, after trying for weeks to even get a meaningful response from them. The general feeling is that no meaningful response can come that doesn't either admit gross negligence, or outright lie (either of which could potentially lose them the case without any further ado). Fullerton's post rests now in the negative hundreds.

Meanwhile, Chipps came across a journal article claiming evidence that Stack Overflow is sexist, because women tend to write questions more than answers. I haven't studied the article's data, so whether that claim is either factually accurate or truly meritous is a conversation for another day.

In response to this discovery, the original notion that answer upvotes should be worth more than question upvotes was turned on its head. The evidence for this tale comes only from a now-deleted post on Meta that was, admittedly, an unauthorised leak of information from a private chatroom. But it's notable that said post has been redacted using the "personal information redaction" feature (it contained no personal information), and no utterance of this rationale has ever been mentioned by staff again, leaving only friendly buzzwords about rewarding contributions (which might have been done better another way, had that really been the goal).

After being told by the small subset of moderators consulted (not all moderators, despite claims by the company) that this would be unpopular, the change was nonetheless performed then revealed in a blog post, triggering predictable negative reactions on meta (where the wider community hadn't been consulted), and sadly caused some long-standing productive contributors to miss milestones for which they'd worked so hard.

It's amusing that the final line of the blog post reads:

These changes [..] are an exciting start to working hand in hand with the community to build a better Stack Overflow.

No. No, they're not.

Commenter Erwin Brandstetter said it best:

The change is bad. Pulling it out of a blackbox, without consulting the user base is worse. But recalculating reputation retroactively is a huge break of trust. A falsification of history. Deceit. You can tell I am not happy with this.

You'd think Stack Exchange might have learnt by now. Still, they persisted.


(A future version of this article will be enhanced with some archived screenshots and other illustrations.)