AI chat responses often disappear from easy access because individual messages lack stable addressing and bookmarking. Users report losing time when a previously correct answer cannot be found after scrolling, and when important drafts become unclear after threads grow. Knowledge work depends on being able to revisit specific text quickly, but AI chat products provide no URL for the most valuable response. The underlying design inherits messaging-app architecture where the conversation is addressable and each message is an ephemeral transcript line. This choice is not required by the actual work; it is driven by product shipping speed and downstream UI decisions rather than by how knowledge should be managed.
"“I had the perfect answer it gave me yesterday about my tax setup and now it's just gone. Scrolled forever, can't find it, and there's no way to bookmark anything. Lost an hour of work basically.” It is one of dozens like it. Another from the same week, on a different product: “Why is there no way to save one message?? I asked it to draft something important, came back later, and the whole thread had gotten so long I couldn't tell which version was the good one.”"
"“The most valuable text a knowledge worker reads in a workweek, the response to a question they could not have answered on their own, is the only text in 2026 that has no URL. AI chat inherited a messaging-app architecture in which the conversation is the addressable unit and the individual message is an ephemeral transcript line.”"
"“Engelbart's 1968 NLS could already jump to any individual text element by structural address. Fifty-eight years later, AI chat cannot.”"
"“AI chat inherited a messaging-app architecture in which the conversation is the addressable unit and the individual message is an ephemeral transcript line. That decision is not load-bearing. It is downstream of how fast the products shipped, not of how the underlying work actually behaves.”"
Read at Medium
Unable to calculate read time
Collection
[
|
...
]