Stack Overflow - Tumblr Posts
Most people wouldn't say that Stack Overflow is in any way obscure, but then again most people don't really go below the surface of simply asking and maybe answering. Most people have no idea that, for example, Stack Overflow has it's own chat server which, although it hasn't been updated in years, is quite possibly the best system of it's type that I've ever used. In fact, there are actually two different servers, one for Stack Overflow and one for the rest of the Stack Exchange Network.
But that's how I got my start to being online. Way back when I didn't even have any social media, I made a Stack Overflow account to ask a terrible (now deleted (not by me (pretty sure it got roomba'd))) question about some especially tricky programming question that I couldn't solve after days of searching. As I used the site more, I began learning more about how it functions and the systems that underpin it, and eventually found my way to meta and then to chat, where I found a very friendly room of people who were the first real software engineers I got to know.
The only problem is that the userbase on stack overflow skews pretty old, especially compared to me at the time, and I think I picked up a lot of my internet-speak from them. Even now, when talking to closer to me in age, I frequently get called out for sounding old or being a boomer, which I think is on account of the dialect of internet I first picked up.
Also, feel like I gotta throw this in, but I know stack overflow gets a lot of shit thrown at it for being hard to approach and for being rude/unwelcoming/mean/heavy handed etc etc I'm sure you've heard the refrain. But as someone who's gotten to know how all these systems work, I promise it isn't all that scary once you understand how to approach it. Anyways, I'm getting off topic, but maybe I'll post about how to be a new user on stack overflow one of these days :P
What was your obscure chatroom/forum that you used before social media? Don’t say Gaia Online or Club Penguin. I’m talking obscure.
big things that are also not good are happening at stack overflow soon :c
Stack Overflow. If you've ever tried writing code, you've seen this website. It's inevitably at or near the top of search results for any programming related query. And, if you've seen anyone on the internet talking about it, you've also probably seen many, many, many complaints about toxicity, gatekeeping, deletion, moderator abuse and, more generally, how asking a question on Stack Overflow is something to avoid doing at all costs.
So, before I talk about the upcoming "bad things" I mentioned yesterday, I'd like to give a little context regarding how Stack Overflow works and why it's so useful despite so many people having terrible experiences.
Stack Overflow is an incredibly effective machine for connecting specific, detailed, well-researched questions to experienced experts who want to answer questions. The key to its success lies in its ability to filter the deluge of incoming questions (nearly one question every second) into a list of questions that would be worthwhile and interesting for someone who wants to answer, and particularly questions that would be applicable to more than just the OP (i.e. ones that you'd find in search results).
This essential filtering process results in the mission/philosophy being somewhat counter-intuitive if you're not already familiar with the site. Rather than a help-desk like focus on getting peoples questions answered, it sounds instead much more like building a community encyclopedia in Q&A form: "We're working together to build a library of detailed, high-quality answers to every question about programming."
How does this filtering work? The answer lies in a shockingly complex and sophisticated curation/moderation system entirely run by ordinary users, supervised by a very small number of elected (unpaid) moderators (~20) acting as "human exception handlers." New questions are scanned by bots, voted up or down, flagged, closed, piped into queues for manual review, edited, reopened, closed again, deleted, undeleted, locked, protected, migrated, you get the idea. [Drop me an ask or a message if you want to know more.]
With this very low bar of entry for doing moderation/curation tasks, there has to be some way to keep everyone on the same page. That mechanism is the "meta" sites -- sites for discussion of what happens on the main sites. For stack overflow that would be meta.stackoverflow.com (and meta.stackexchange.com for cross-site issues). Over the years, discussions and questions on the meta sites have resulted in the policies and guidance that users are expected follow today. Importantly, nearly all these policies were created, revised, and maintained by the community and volunteer moderators over the lifetime of the site.
Some years ago, the company and the community felt in-tune. Employees would frequently talk on the meta sites about what they were working on, and the employees on the community management team were highly respected, especially by the volunteer moderators. Feature requests posted on the meta sites would actually get responded to and sometimes implemented. But in recent years that has largely deteriorated. Major (poorly handled) firings of beloved community managers, strange, half-baked projects being worked on and then abandoned, a sharp decrease in communication and openness with the community.
Specifically of note to those who have heard/seen that Stack Overflow is toxic was a push from the company to make the site more welcoming, which, rather than being met with opposition, was met with a surge of ideas and brainstorming from community members who also recognize that the site does not provide anywhere near enough support to new users. All of this community brainstorming was unfortunately ignored. They ended up replacing the "Be Nice" policy with a proper code of conduct and added a little waving hand next to new-users names though, so problem solved! /s
This finally brings me to what has prompted me to write this post. Over the last few years, the company has made blunder after blunder, but largely in ways that leave the site in the same state as it was before with small UI changes or minor tweaks. The community chugs on, although they are increasingly disappointed in the state of their beloved website. But this recent change is the first one that poses an existential threat to Stack Overflow as we know it.
When ChatGPT and similar LLMs first began to hit the mainstream months ago, they were prompted banned on Stack Overflow. Say what you will about how good LLMs are, but this policy was instantiated by the Stack Overflow community moderators after seeing shocking numbers of un-reviewed, copy pasted, incorrect and hallucinated answers being posted by users playing with their new toys. The ban has stood for 6 months and served the site pretty well. In theory there is room to use AI in answering questions, but the frequency of correct-looking answers with subtle errors makes it time-consuming enough to fact check that it's worth the potential losses to delete everything with the hallmarks of LLM usage.
However, the stack overflow community management team has recently delivered a new policy. This policy deviates from the extensive history of trusting community moderators to use their best judgement and do what their community asks of them. It instead prescribes that community moderators apply "very strict standard of evidence" when determining if a post was GPT generated or not. The specific criterion that moderators were instructed to use are non-public so I can't share them, but I can say that it's restrictive to the extent that it will be effectively impossible to delete GPT generated content. This is the first time in my memory that the community moderators have ever been forced to implement a site policy by the company. It's a huge blow to the community driven nature of the sites, and, once the policy is live, it will slowly degrade the usefulness of the site. By making Stack Overflow no better than plugging a question into ChatGPT, there will be no reason to use it. It's an absolutely terrible policy for a vast array of reasons, and even after strong pushback by community members and moderators, the company is not backing down.
In response to these changes, there's a moderation strike being coordinated by multiple key contingents of the userbase (moderators, anti-spam teams, user groups for curation) to see if we can get the company to change course. That strike is set to begin on Monday, and I'll post some details once that goes live.
I hope that regardless of how you feel about Stack Overflow or AI generated content, you can see how unprecedented these changes are and what a last-resort measure this moderation strike is, and I hope that you can support us :]
Or, short version:
ChatGPT comes out. People start copy pasting ChatGPT answers onto stack overflow without verifying them. Policy is created to ban ChatGPT content.
6 months pass. Moderators figure out sophisticated ways to detect GPT content. Because the text is designed to sound humanlike, this typically involves factors outside of the content of the answer, in addition to in-answer stuff (like it ending with "Regenerate response" or calls to non-existant APIs).
Company comes out of nowhere and says publicly "Golly gee our mods have been sticking posts into shitty AI detectors and if the AI detectors (which we all know are garbage) say it's AI, they delete it. So we're telling the mods to stop doing that! They can still ban GPT, but just not based on detectors :)" while privately saying "Stop deleting GPT content outside of [very very narrow edge case which I am prohibited from sharing due to the Mod Agreement]." This edge case is so narrow that nearly nothing falls into it btw! Have fun explaining to the users why their flags are being declined ^.^
Mods: "what"
Company: "also your deletions target certain geographic regions more than others. so stop being racist."
Mods: "what. that's. that's not how any of this works. spammers use GPT in spam and spam comes primarily from Asia / India. so of course the deletions are geographically biased. HUH???"
So, strike started today, involving a huge number of power users and moderators. The article cites 11% of moderators, but that's a bit of a weird number. Many of the smaller *.stackexchange.com sites haven't signed on because it's not a problem over there. Of the stackoverflow.com mods, well, ~75% (15 out of twenty something) are officially on strike and the remainder are not doing much. Also notably, stack exchange is heavily reliant on a community project (Charcoal SE / SmokeDetector) for spam detection, a project which has... shut down for the duration of this strike. Here's the homepage of askubuntu (a stackexchange site) (not my screenshot so don't go hunting down the account):
Relevant links (I'll try and post when there's new stuff):
Open letter drafted by the community: https://openletter.mousetail.nl/
Original GPT ban on stack overflow: https://meta.stackoverflow.com/q/421831
Public facing new policy: https://meta.stackexchange.com/q/389582
Moderator response to questions from users regarding the new policy: https://meta.stackoverflow.com/a/424911
Good summary of the events leading up to the strike from a Stack Overflow Mod: https://meta.stackoverflow.com/q/424979
Some insights from a widely respected former Community Manager (company employee): https://jlericson.com/2023/05/31/mod_strike.html https://jlericson.com/2023/06/02/strike_comentary.html
Strike announcement: https://meta.stackexchange.com/q/389811
Clarification regarding some misconceptions about the usage of GPT detectors: https://meta.stackexchange.com/q/389828
Company statement responding to the strike: https://meta.stackexchange.com/q/389834
Quick little stack overflow mod strike update: The company released a much more detailed (if flawed) analysis from an internal data scientist which they interpreted as pointing towards the GPT ban being the cause of low user retention. This analysis, while kinda neat, does not really actually show anything. It does however explain a little bit more of how we got here.
(And, for the first time in a long while, is a substantive post by a staff member that didn't get downvoted to hell)
Week 2: Stack Exchange mod/curator strike
Negotiations begin
The company has asked for a small (3 person) delegation to discuss issues with and try to reach a mutually agreeable end to the strike. These three members have been selected by vote and are all highly respected and capable moderators. Talks are expected to begin today. Unsure where that'll go, since the demands were laid out pretty clearly in both the strike letter and on meta, but at least there will be some kind of communication.
Stack exchange refuses to interact directly with the community, which, while on the one hand would be challenging in this situation due to the size of the community vs the size of the staff, on the other hand the lack of communication with the community is exactly what got us into this mess so we'll see how it goes.
The case of the missing data dump
To throw fuel on the fire, Stack Exchange announced late last week that they also intend to do as twitter and now reddit have done -- close up access to their API. But rather than announce it, they implemented their first step without telling anyone, over a month ago. That first step was turning off the public data dumps that stack overflow has been publishing once every 3 months since 2014. The community noticed that the newest one hadn't gone up and asked about it, receiving a response from a former dev and then from the CTO explaining that not only was the dump gone, the API and data explorers were also going to be locked down.
Aaaaaand of course the moment I post this, a former community manager shares a new blog post they've written, containing leaked internal comms from Stack Exchange. Nothing terribly new or surprising, but it's an interesting insight:
Week 2: Stack Exchange mod/curator strike
Negotiations begin
The company has asked for a small (3 person) delegation to discuss issues with and try to reach a mutually agreeable end to the strike. These three members have been selected by vote and are all highly respected and capable moderators. Talks are expected to begin today. Unsure where that'll go, since the demands were laid out pretty clearly in both the strike letter and on meta, but at least there will be some kind of communication.
Stack exchange refuses to interact directly with the community, which, while on the one hand would be challenging in this situation due to the size of the community vs the size of the staff, on the other hand the lack of communication with the community is exactly what got us into this mess so we'll see how it goes.
The case of the missing data dump
To throw fuel on the fire, Stack Exchange announced late last week that they also intend to do as twitter and now reddit have done -- close up access to their API. But rather than announce it, they implemented their first step without telling anyone, over a month ago. That first step was turning off the public data dumps that stack overflow has been publishing once every 3 months since 2014. The community noticed that the newest one hadn't gone up and asked about it, receiving a response from a former dev and then from the CTO explaining that not only was the dump gone, the API and data explorers were also going to be locked down.
While I understand that Stack Overflow inherently just... has a smaller audience, it's crazy to me how little attention it gets over doing shit that's so much worse than putting a price tag on their API. Probably also doesn't help that everyone and their manager thinks it's a toxic shithole but people think that about reddit too so idk.
So like, the Reddit strike going on right now, yeah? I've been seeing a lot of people comment on how they appreciate the protest and then go on to say that this has the notable downside of them constantly looking up questions and not being able to easily find the answers because all of the easily-findable answers are exclusively on Reddit. I am not sure if most of the people making this observation are within the line of thought of "man, maybe this protest isn't such a good idea after all" or "man, it really sucks that we've let the internet get so consolidated," and I'm really hoping its the latter.
Like, all of this? This right here? Reddit making a shitty, anti-consumer grab for money and control over how people are allowed to access the information on their servers, and the website going dark in protest causing tons of people to not be able to access important information? This is exactly what people mean when they say that it's bad that the internet has shrunk down so much and is mostly comprised of, like, 10 websites. It's a fucking problem that one company making one bad decision and causing their website to crash and burn can jeopardize so much of humanity's cumulative information.
This two-day glimpse into the internet without Reddit is the warning shot. Imagine what will happen if Reddit actually goes down for good for one reason or another one day. Imagine what will happen if/when Discord or Fandom bites the dust, or gets rendered practically-unusable without paying an ever-increasing premium because they're owned by blood-sucking corporate leeches.
Another big thing is Twitter clamping down really hard on your ability to DM people if you don't have Twitter Blue. If this goes through, it'll put a ton of artists and sex workers who rely on Twitter DMs for their business operation into a shitty situation. Now, obviously, it's not gonna be the end of the world for them, but once again, it feels like a warning shot to me. Twitter is a sinking ship, and unless something changes and it starts to course-correct, I worry that it'll go under and all of the creators who rely on it will suddenly be in an extremely precarious situation.
These are the sorts of things that we, as the users of the internet, need to seriously think about as time goes on, and if we don't find an adequate answer sooner, we're going to pay for it later. I still hold that the best solution is to start making and using more individual, niche websites. Things like Twitter, Reddit, Discord, etc. have their place, of course, but I seriously think a lot was lost through the death of things like individual forums and the existence of many different wiki-hosting sites.
We need a concerted effort, not just on the side of larger creators, but on the users themselves, to stop exclusively using these larger websites and support the creation and growth of smaller, more niche websites, and prevent a catastrophe before it actually happens. I simply hope that people with larger platforms than my own pick up on all this and start talking about it and swaying people to act sooner rather than later. I know it's possible to correct the problem of the mysteriously tiny internet before a modern Library of Alexandria moment happens, I just don't know if that correction will actually happen in time.
more detail tomorrow but wow a lot of parts of Stack Overflow Incs positions on the mod strike have just started to make a LOT more sense in the stupidest way possible
Hi I forgot I promised this but here it is:
So tl;dr for those not up to date, Stack Overflow has a longstanding community-created policy banning GPT content, then out of nowhere the company effectively told moderators that they can no longer delete any GPT content. Very confusing move and nobody really knew why until...
They announced a new feature for the stack overflow editor! It was pitched as a sort-of Grammerly + code linting combo under the name "AI Formatting Assistant" (because everything has to be AI these days) to help the new user experience. The idea was that if the site could help people ask better questions, they'll get better answers and have a better experience with the site! I personally thought this was a great idea, and was excited to see the rollout.
The reality of this feature was... well you can see for yourself. To save you a click, it would take the question and send it directly to ChatGPT, prefixed with a prompt (which was broken out of within 5 minutes of release). It was horribly broken in many ways, changing code, changing meaningful parts of questions, all of the things that you'd expect for a generative tool on a task that is not generative in nature.
However this seems to be a particularly interesting feature with the context that they had recently stopped moderators from moderating GPT content and I think it's pretty easy to see the connection :P
more detail tomorrow but wow a lot of parts of Stack Overflow Incs positions on the mod strike have just started to make a LOT more sense in the stupidest way possible
every time i see a nother website implement useles ai i just feel soooo so so so so so. sososo so so so soo sososo.
Neat article that sums up the stack overflow, reddit, and wikipedia struggles with generative AI pretty nicely. I'm not an AI expert but I have been witnessing this over at Stack Overflow, and I thought this was pretty solid.
This is a post about the recent changes announced on @staff earlier today. I drafted this story for other purposes but I think it's important enough to warrant posting here as well:
While I am still relatively new to tumblr (only a little over a year old) I have just gotten through with watching another site I care deeply about wander off a cliff while desperately searching for profitability and I am deeply concerned that tumblr is about to set course along a similar path.
Before I go into detail about the case study that's motivated me to write this advice, here's the gist of it: Keeping your existing power users, resolving their complaints and empowering them to do more will get you leaps and bounds further than chasing new users and new pageviews.
The case study is a site I'm sure everyone here as heard of: Stack Overflow. While we all have come to know it for its reliability and prevalence in search results for a huge swath of programming related queries, that site, like this one, is driven by user generated content and user curation. The reason why Stack Overflow was able to achieve what is has today is due to a combination of very good systems for curating content (both that obvious part of ensuring answers are correct, and the much less obvious part of ensuring that high quality questions are easy to find, an essential part of attracting and retaining expert answerers) -- to put this in simple terms, the site is able to do what it does due to a (very small) group of power users. I don't have your internal statistics, but I would wager that tumblr has a much larger percentage of "power users" than Stack Overflow does, which is something to be thankful for.
Around 2016, Stack Overflow realized it had a problem. Well, two problems: First, the number of new questions and answers coming in were falling and second, they had been in business for a decade and needed to start making money. To do this, they shifted their focus from their main sites (referred to as "Public Platform") to other initiatives, marketing their enterprise offering more aggressively and building Stack Overflow for Teams. In those years, they launched several failed expansions to the public platform (most notably Documentation). As a user, I was glad to see them doing this -- I wanted them to be profitable! I need them to be profitable in order to guarantee their long term existence.
The decreased focus in public platform did hurt, the pace of new features in that area slowed to a crawl and many community members didn't feel particularly heard during those years, but the site still worked. Leaving things the same doesn't tend to chase already heavily invested users away. However, in mid 2018 focus returned to the public platform and the goal was to make it profitable by itself. The path to profitability that the company saw was through getting more users participating, and what better way to do that than to attempt to convert more passive viewers to active users. So, why aren't users converting?
Stack Overflow had a new user experience problem, one far worse than tumblr. I have seen numerous instances of advice to never post on Stack Overflow due to the toxicity that you were all but guaranteed to receive in return. The community of power users was also keenly aware of this problem -- nobody wants to live on a site that the world labels as toxic. The (approximate) consensus among the power user community was that the reason why new users were having such a negative experience was due to mismatched expectations about what the site should be. New users treated the site like a helpdesk or forum, while power users treated the site as a highly curated knowledge base. This clash in expectations around quality resulted in a terrible new user experience, exacerbated by the fact that the UI of the site did a terrible job of communicating what the expectations were. I believe that tumblr also suffers from a similar (but far less extreme) clash of expectations -- an existing userbase that wants a site that works in very specific ways and new users coming from sites that function very differently.
To resolve this clash, in 2018 Stack Overflow launched what they called the "Welcoming Wagon" -- a series of efforts aimed at making the site friendlier to new users, meeting new users where they were rather than guiding new users to produce better content. In their view, the issues with the site were not the fault of new users, it was the fault of *existing users* for not providing a welcoming environment for new users. So, while they realized that these changes would most likely alienate some of their core userbase, it was justified as worthwhile because the power user userbase was a tiny fraction of the total userbase (in a now notorious post, an employee pointed to the statistic that power users were less than one hundredth of a percent of active users!). Wouldn't losing a few members of that group be worth the gains?
This marked a turning point in the companies history, and rather than diving into each and every change that was negative I will point out the 2 widely appreciated features that did make it through, not because they are directly translatable to tumblr but because the illustrate that new ways in which the company worked *with* the community to great effect: First, a totally new look at their question editor, prompted by years of community feedback, and created in collaboration with community members, to help guide new users into asking questions that are a good fit for the site. Through discussion and feedback from the community, this editor was able to notably improve new question reception. And second, two iterations of mentorship programs which allowed new users to receive targeted feedback and work alongside power users to improve their questions before they were posted to the public site, which also resulted in those users encountering more success.
However despite these two spots of light in the darkness, the total lack of improvement in power user tooling while also blaming power users for the sites issues has slowly prompted its curators and answers leave the site over the last 5 years, and the most recent action (allowing GPT generated content) has culminated in strike by many moderators and power users which has recently reached its 1 month mark. The VP of community team is currently negotiating with a few moderator representatives, and there does appear to be some chance of resolution, but to me the strike sounds like a bell tolling for the death of Stack Overflow. Of course, a site like Stack Overflow cannot die in a day and will probably continue to chug along for a few years yet, but without its answers and curators the site will eventually fall into obscurity -- many answers already suffer from being outdated, and the rates of spam have skyrocketed. On many days this month the front page has been entirely spam and bots, and inevitably people will start running into that from search engines.
All that to say, there are many good things for tumblr in the recent post from @staff. SEO has been a pain point for years. So has search. The performance and stability of the apps is clearly lacking. And I certainly am glad they are looking to convert more users -- that really does benefit everyone, regardless of how people gripe about the users coming from <xyz> site being terrible. But I sincerely hope that they can learn from Stack Overflow's example and *teach new users how the site works* rather than focusing on *changing how the site works to make things easier for new users.*
Looks like the Stack Overflow strike is finally coming to a close, after nearly 8 weeks, having achieved all of its major goals (with only a few, small, compromises)! Just wanted to put something out since I'm pretty excited about it. Once it's official and the meta post is out I'll see if I can put together a more detailed analysis of negotiations and why this went so differently from other strikes such as the one on reddit.
Here's the official post with the results of the strike negotiations if you really like reading, but I'll try and put out a more layperson friendly version in the coming days.
Looks like the Stack Overflow strike is finally coming to a close, after nearly 8 weeks, having achieved all of its major goals (with only a few, small, compromises)! Just wanted to put something out since I'm pretty excited about it. Once it's official and the meta post is out I'll see if I can put together a more detailed analysis of negotiations and why this went so differently from other strikes such as the one on reddit.
The end of the Stack Overflow Strike
Note: Skip to the bottom if you just want a quick list of strike outcomes and not story mode.
For those of you who missed it (I know many of the engineers I talk with day to day did), starting on June 5th the Stack Exchange elected moderators and many curators and users announced they were on strike. The relationship between the company and the community had been slowly slipping for the last 5+ years as company actions gradually degraded the once robust trust that community members had in how their platform was being run. However, none of these changes had truly threatened the long term existence of the site like the decision to unilaterally override the existing, community established policy to ban GenAI content with a new, secret, moderator only policy that effectively forced moderators to stop moderating GenAI content (this policy as since been released, and I'll link it later).
The combination of a policy whose contents were secret and which went directly against community consensus (and which many believed would ultimately result in the death of the site) convinced over 100 moderators and over 1500 users to sign on to the strike letter. These users included two key groups: first, Charcoal, a highly effective community built and maintained spam prevention group, chose to shut down their operations as a part of the strike, and second, the Stack Overflow moderator team agreed to stopped moderating, resulting in a backlog of nearly 20,000 unhandled flags during the strike.
This impact was enough to bring the company to the table. Despite despite banning strike organization from taking place on stack exchange chat rooms or sites (which required moving to a discord server for organization) and hush-hush messages to striking moderators implying that their moderator status may be revoked under certain conditions, the community held strong and the strike began.
Mere days after the strike began, the VP of Community reached out to ask for a set of community representatives to negotiate with in private. Elections ran through week 1 of the strike and ultimately 3 moderators were selected who would communicate with strikers in the strike discord and communicate with the company representatives in private, allowing negotiations began at the start of week two. These negotiations were interrupted by a couple of events which I've previously written about:
First, the botched release of the "AI Question Assistant" which turned out to pass questions to the ChatGPT API with a prompt and return the result (quickly jailbroken, widely mocked, and ultimately scrapped by the company). It had problem such as changing what question was being asked, fixing bugs in code, and answering the question instead of fixing it.
Second, the removal of the Data Dumps (with the implication that this was to prevent GenAI companies from training on it), which turned out to be a directive straight from the CEO and was quickly reverted after a call from one of the company's founders.
And finally, clear misrepresentations of the strike to any media outlet that contact them.
After the first month, spirits were not great. Progress was occurring but at nowhere near the pace that people had hoped for. The strikers had hoped for an immediate retraction of the GenAI policy, which had not yet happened, however what the negotiators were able to share sounded promising, so the strike continued.
Finally, after nearly 8 weeks on strike, the negotiators came to an agreement and announced that Stack Overflow had conceded on nearly every concern. Here's a quick rundown of the results:
One strike demand was that the private policy moderators had received would be published and so it was. This made it clear to everyone how damning that policy was. In summary, it restricted moderators to only act on self-admission and otherwise treat GenAI content as human (even when it includes clear tells such as talking about a knowledge cutoff date).
The second was that this moderator-only policy was retracted. It was replaced with a heuristic based policy that would allow moderators to act on certain heuristics, which was effectively the status quo prior to the strike.
They also made a commitment to keep the data explorer, the data dumps, and the API freely available. This wasn't initially part of the strike demands but was added after the data dump snafu in the middle of the strike.
They agreed to never mandate moderator actions in private that were not justified by a public policy.
They agreed to update their press policy to ensure that statements about the community are reviewed by their internal community team and not sent out without consultation (which is apparently what happened).
They granted moderators an ability to establish when the company has failed to hold up its end of the moderator agreement (by vote). If the moderators established that the agreement was violated, they have committed to reverting any changes that were made improperly and do them again, in a way that is compliant with the agreement.
They updated their policies around how they will handle support cases complaining about moderator action to involve actually consulting with the moderator(s) who made the decision, and also their policies around communicating with moderators more widely. Mods have been asking for a return to more active and open communication from the company and these changes are a huge step in the right direction there.
And finally they made some softer commitments around taking community feedback under account, being more transparent with platform changes, and making an effort to communicate more with users and moderators.
These changes are massive and while there is still some talk of continuing the strike until there is more follow through on these commitments, many users are chomping at the bit to get back to work. The vote currently stands at around 95% approval for ending the strike, so I think we can safely call this over, and a massive win for the Stack Overflow community. And if there's more you want to know about Stack Exchange or the strike itself, my askbox is open ^.^
Little update, within 24 hours of the end of the strike the stack overflow mods handled over 10k of the collected flags. Those folks are machines and are incredibly good at what they do, illustrating exactly how much the company needs them.
The end of the Stack Overflow Strike
Note: Skip to the bottom if you just want a quick list of strike outcomes and not story mode.
For those of you who missed it (I know many of the engineers I talk with day to day did), starting on June 5th the Stack Exchange elected moderators and many curators and users announced they were on strike. The relationship between the company and the community had been slowly slipping for the last 5+ years as company actions gradually degraded the once robust trust that community members had in how their platform was being run. However, none of these changes had truly threatened the long term existence of the site like the decision to unilaterally override the existing, community established policy to ban GenAI content with a new, secret, moderator only policy that effectively forced moderators to stop moderating GenAI content (this policy as since been released, and I'll link it later).
The combination of a policy whose contents were secret and which went directly against community consensus (and which many believed would ultimately result in the death of the site) convinced over 100 moderators and over 1500 users to sign on to the strike letter. These users included two key groups: first, Charcoal, a highly effective community built and maintained spam prevention group, chose to shut down their operations as a part of the strike, and second, the Stack Overflow moderator team agreed to stopped moderating, resulting in a backlog of nearly 20,000 unhandled flags during the strike.
This impact was enough to bring the company to the table. Despite despite banning strike organization from taking place on stack exchange chat rooms or sites (which required moving to a discord server for organization) and hush-hush messages to striking moderators implying that their moderator status may be revoked under certain conditions, the community held strong and the strike began.
Mere days after the strike began, the VP of Community reached out to ask for a set of community representatives to negotiate with in private. Elections ran through week 1 of the strike and ultimately 3 moderators were selected who would communicate with strikers in the strike discord and communicate with the company representatives in private, allowing negotiations began at the start of week two. These negotiations were interrupted by a couple of events which I've previously written about:
First, the botched release of the "AI Question Assistant" which turned out to pass questions to the ChatGPT API with a prompt and return the result (quickly jailbroken, widely mocked, and ultimately scrapped by the company). It had problem such as changing what question was being asked, fixing bugs in code, and answering the question instead of fixing it.
Second, the removal of the Data Dumps (with the implication that this was to prevent GenAI companies from training on it), which turned out to be a directive straight from the CEO and was quickly reverted after a call from one of the company's founders.
And finally, clear misrepresentations of the strike to any media outlet that contact them.
After the first month, spirits were not great. Progress was occurring but at nowhere near the pace that people had hoped for. The strikers had hoped for an immediate retraction of the GenAI policy, which had not yet happened, however what the negotiators were able to share sounded promising, so the strike continued.
Finally, after nearly 8 weeks on strike, the negotiators came to an agreement and announced that Stack Overflow had conceded on nearly every concern. Here's a quick rundown of the results:
One strike demand was that the private policy moderators had received would be published and so it was. This made it clear to everyone how damning that policy was. In summary, it restricted moderators to only act on self-admission and otherwise treat GenAI content as human (even when it includes clear tells such as talking about a knowledge cutoff date).
The second was that this moderator-only policy was retracted. It was replaced with a heuristic based policy that would allow moderators to act on certain heuristics, which was effectively the status quo prior to the strike.
They also made a commitment to keep the data explorer, the data dumps, and the API freely available. This wasn't initially part of the strike demands but was added after the data dump snafu in the middle of the strike.
They agreed to never mandate moderator actions in private that were not justified by a public policy.
They agreed to update their press policy to ensure that statements about the community are reviewed by their internal community team and not sent out without consultation (which is apparently what happened).
They granted moderators an ability to establish when the company has failed to hold up its end of the moderator agreement (by vote). If the moderators established that the agreement was violated, they have committed to reverting any changes that were made improperly and do them again, in a way that is compliant with the agreement.
They updated their policies around how they will handle support cases complaining about moderator action to involve actually consulting with the moderator(s) who made the decision, and also their policies around communicating with moderators more widely. Mods have been asking for a return to more active and open communication from the company and these changes are a huge step in the right direction there.
And finally they made some softer commitments around taking community feedback under account, being more transparent with platform changes, and making an effort to communicate more with users and moderators.
These changes are massive and while there is still some talk of continuing the strike until there is more follow through on these commitments, many users are chomping at the bit to get back to work. The vote currently stands at around 95% approval for ending the strike, so I think we can safely call this over, and a massive win for the Stack Overflow community. And if there's more you want to know about Stack Exchange or the strike itself, my askbox is open ^.^