Group chats are the new forums, and the archives are worse

Open a Discord server for any contemporary open-source project. Twenty people are talking about how to make the library work with a specific edge case. Half of them have hit the same problem before. The other half are about to. The first half know things the second half need. None of it is searchable from outside the server, none of it appears in any search engine, and three weeks from now nobody, including the people in the conversation, will be able to find it again.

This is the conversation that used to happen on a forum. The same people, the same problem, the same answer. The difference is that the forum thread was a URL. It was indexed. It was searchable. A year later someone with the same problem would find the same thread, and the answer would still be there, with replies, with corrections, with the person who solved it linked to other things they had solved.

That archive has stopped being written.

What Stack Overflow's traffic actually says

The clearest evidence runs through Stack Overflow. Through the 2010s, Stack Overflow was the de facto search engine for technical questions. For most routine problems a working engineer hit, there was a Stack Overflow answer with the question above it, the explanation below, and a comment thread with the corrections. The archive compounded for fifteen years.

Then two things happened. AI assistants started giving the same answers without making the user click a link. And the conversation that produced the next generation of answers moved off Stack Overflow and into Discord.

Public traffic estimates for stackoverflow.com show a roughly 50% decline between 2022 and 2024. Stack Exchange's own data dumps tell the same story from the supply side: question volume has dropped sharply, and the questions that still get posted often go unanswered. The answer-velocity that fed the archive has slowed to a trickle.

A UI change is not automatically a loss. If AI assistants give correct answers and users solve their problems, replacing Stack Overflow as a destination costs us nothing. The problem is what happens next: the assistants are trained on Stack Overflow's archive. When the archive stops being updated by humans, the assistants should be expected to hallucinate more on new edge cases, because the training data ages out. Whether that is already happening at scale is not something we have seen measured.

Human-written, search-indexed technical Q&A is no longer being produced at the rate it used to be. The model that fed it has broken.

Why Discord is the wrong replacement

Discord is a wonderful chat product. It is a terrible knowledge base.

The search inside a Discord server can only find messages in the channel you are currently in, with several documented constraints around date range and term length. There is no cross-server search. The full-text index is, by the company's own documentation, optimised for finding a recent message you remember posting, not for surfacing a question someone else asked two years ago.

Search engines cannot index Discord conversations. The content lives behind authentication. The Discord crawler exclusion is not even an active choice; it is what happens when a platform is structured as a closed account-required chat by default.

When a question gets asked and answered in a Discord server, the answer exists. It is read by the people in the channel at the time. It is also, in nearly every meaningful sense, gone. A new person with the same question, two weeks later, will ask the same question again. The same expert will type the same answer. The cycle repeats forever because nothing about Discord's design rewards turning chat into archive.

This is not a Discord-specific problem. Slack has it. WhatsApp groups have it. Group iMessage threads have it. Every modern team chat product is built around real-time communication, not searchable retention. The shift away from forums is a shift across the entire category.

What forums got right

The phpBB-era forum had three properties that, in retrospect, were extraordinary.

It had a URL for every thread. A specific URL, that could be linked, shared, bookmarked. The link survived browser changes, account changes, decade changes. You could send someone a forum thread URL in 2024 and they could read the discussion from 2007 with no friction.

It had search that Google could see. The content was public by default. Google indexed it. A person searching for a problem could find the right thread, read the answer, and move on. The cost of asking the same question twice was high because the first asker's thread was findable. So new questions tended to be new.

It had a moderation model where a community grew up around topic experts who answered questions they cared about, in public, where everyone could see the answer was correct. The reward for the expert was reputation that compounded. The reward for the asker was a definitive answer they could link back to.

None of Discord, Reddit, or a shrinking Stack Overflow now delivers all three.

The infrastructure of public Q&A is, in 2026, in worse shape than it was in 2016.

Where the knowledge actually goes now

It does not just disappear. It moves. We can trace where.

Some of it migrates to YouTube, where individuals answer common problems in long-form videos. The medium is better than nothing. It is also worse than text for skimming, searching, copying code from, and updating when the answer changes.

Some of it migrates to private newsletters. Specific experts in specific domains have substantive Substacks that the audience pays for. The content is good. It is also paywalled and the long-term archive depends on the writer continuing to maintain hosting.

Some of it stays in Discord and dies there. Anecdotally this feels like the biggest share, though nobody has measured the split. The expert who would have written a blog post in 2014 now answers in a Discord channel, helps three people, and the answer has the lifespan of the scrollback buffer.

Some of it goes into private GitHub repos as README updates, internal company wikis, and Notion pages that the next employee won't be able to find when the person who wrote it leaves.

What would help

If you maintain a community on Discord, consider running a public forum or a public wiki alongside it. Not instead of. Alongside. The same expert who answered the question can spend five minutes pasting the answer into a public, indexed, durable URL. The next person to hit the same problem will find it. The expert will not have to type it again.

This is not a hard ask. It is also, observably, almost nobody's job, and the shift happened before anyone debated whether it should. The window of publicly indexed technical expertise ran roughly from the phpBB era to the mid-2020s. What replaces it is not yet a thing that can be searched.