Digital Lore

Jason Scott, the Internet Archive's resident angel of digital death, has been warning for years about a specific kind of knowledge loss that doesn't look like knowledge loss. It looks like a chat server with 200,000 members.¹

The argument is about "lore" — the informal, contextual, often messy knowledge that exists between formal documentation. It's the difference between the official manual for a synthesizer and the forum post explaining that you need to turn down the cutoff resonance before changing patches or it'll make a horrible noise. Lore is the grease between the concrete blocks of knowledge. It's the workaround for the undocumented bug, the trick passed between practitioners, the answer to the question nobody thought to put in the FAQ.

The Platform Dependency Problem

Discord has become the primary vessel for lore in enormous numbers of communities — hobbyist, professional, technical, creative. This happened not because Discord is good at preserving lore, but because it's good at generating conversation. People flock there because it's easy, real-time, and social. The lore accumulates as a side effect of socializing.

The problem: Discord is a private company with no obligation to preserve anything. It has no export feature worth mentioning. Its search is mediocre. Conversations scroll past and are never seen again. The lore is there — buried in thousands of messages across hundreds of channels — but it's effectively inaccessible to anyone who didn't happen to be present when the relevant conversation occurred.

Scott draws the comparison to ImageShack, the free image hosting service that, in a pivot to subscription pricing, deleted thirteen years of legacy images. This wasn't just a loss of funny pictures. ImageShack had become the unofficial image host for thousands of technical forums. Wiring diagrams, circuit schematics, step-by-step repair guides — all rendered as broken image links overnight. Walking through the affected forums was "walking through a bombed city, its institutional and cultural memory pockmarked with 'pay us to see this stuff' placeholders."

Discord will do this too. Not necessarily through malice, but through the ordinary lifecycle of venture-backed companies: growth, monetization pressure, pivot, enshittification, death. OpenFeint — the gaming social platform that Discord's founders previously built — went from founding to acquisition to shutdown in three years. Discord has lasted longer, but the structural incentives are the same.

Lore vs. Knowledge

The distinction between lore and formal knowledge matters because the transfer from one to the other is lossy, messy, and often doesn't happen at all. The ideal case: someone with deep practical experience writes a comprehensive guide, and the lore becomes documentation. But this requires someone motivated enough to do the thankless work of systematizing what practitioners know implicitly. Most lore never gets this treatment. It lives and dies in the medium where it was generated — the forum, the chat room, the mailing list.

Stack Overflow partially addressed this for programming: the Q&A format forces lore into a structured, searchable form. But Stack Overflow's model doesn't generalize well to domains where questions don't have single correct answers, or where the useful knowledge is contextual and conversational rather than factual and discrete. And even Stack Overflow is a centralized platform subject to the same lifecycle pressures as Discord.

This connects to lost and found knowledge in a specific way. The historical pattern of knowledge loss — Roman concrete, Greek fire, the Baghdad House of Wisdom — usually involves physical destruction or civilizational collapse. Digital lore loss is different: the knowledge isn't destroyed by catastrophe but by platform economics. The medium that made the lore easy to generate also makes it fragile. Chat is optimized for real-time conversation, not for preservation. The better the conversation tool, the worse it is as an archive.

The Write-Once Read-Never Problem

Discord messages are, as Scott puts it, "write-once read-never rants and dumb questions, punctuated with someone asking a question for the hundredth time and someone answering a different way." The signal-to-noise ratio is terrible. But the signal is there, and when the platform goes dark, the noise dies alongside it.

Scott's proposed solutions are modest and, as he admits, unlikely to be implemented: a "Lore" channel where moderators can pin important information permanently; an export feature for this channel; some kind of FAQ system. These would help, but they work against Discord's business incentives. Discord wants you in Discord, talking, engaging, boosting metrics. It doesn't want you exporting knowledge to formats that don't require Discord to access.

The deeper issue is that we keep building knowledge infrastructure on platforms whose incentives are orthogonal to knowledge preservation. Forums were better than Discord for this — they were searchable, browsable, and indexed by Google. But forums lost to Discord on ease of use and social dynamics. Wikis are better still, but they require effort to write and maintain — effort that most communities can't sustain. The medium that wins the attention war is rarely the medium that wins the preservation war.

For a knowledge base like this wiki, the lesson is concrete: anything that lives only in a chat platform is on borrowed time. The act of synthesizing lore into permanent written form — messy and lossy as that process is — is the only thing that gives knowledge a chance of surviving the platform that generated it.