The Slack channel that became the wiki

A new engineer joined a series-B startup on a Monday. By Friday, she had been told four times to scroll back in #eng-general for context. She had scrolled back. The channel had three years of messages. She found seventeen partial discussions of the architecture decision she was trying to understand, none of them resolved, two contradicting each other, and one ending with a thread reaction emoji that was apparently the company's way of marking consensus.

That was the wiki. There was a Notion workspace with sixty-two pages, of which forty-one had been updated more than a year ago and twelve had titles that began with [OLD].

The Slack-as-wiki problem is the default state of every company that does not have a real documentation discipline, and most companies do not. Slack is comfortable to write in. Slack rewards writing with reactions in real time. Slack has a beautiful onboarding flow for new conversations. Slack is also the worst possible knowledge base ever invented, because every decision in it scrolls into oblivion at the speed of the next channel post, the search is keyword-based against ephemeral language, and the only indexing is timestamp.

The new hire pays this cost most visibly, but the cost is being paid by every existing employee on a quieter schedule. The architecture decision made in a Slack thread in 2023 is being relitigated in a Slack thread in 2024 because nobody could find the 2023 thread, and the team is making the same call against the same tradeoffs, with the same arguments, for the second or third time. Every quarter of every year. The compounding waste is enormous and invisible because no individual instance feels like waste.

The fix is not a better wiki. Better wikis fail with the same regularity as bad wikis, because the failure mode is incentive, not tooling. The fix is a decision log. Five lines per entry. What we decided. Why. When. Who decided. What would make us revisit it. The log is owned by the decider, not by a documentation team. Entries are appended, not edited. The log is searchable, dated, and small enough that it actually gets written.

The discipline holds because the bar is low. A five-line entry takes ninety seconds. The cost is small enough that the decider does it before the energy fades. The benefit is large enough that any reader, six months later, can reconstruct the context in three minutes. The decision log is the only documentation artifact I have ever seen survive at scale, because it is the only one whose unit cost is lower than the friction of writing.

Slack will continue to be where the discussion happens. That is fine. Slack is good at discussion. The decision that the discussion produced needs to live somewhere that doesn't scroll. The five-line entry is the cheapest version of somewhere. Without it, your wiki is your Slack history, and your Slack history is unfindable, and your new hires will spend their first month doing the same archaeology every previous new hire did.