Somewhere on a lot of people's hard drives sits a Roam Research or Obsidian vault from a few years back — a few hundred notes, a link graph that looked genuinely impressive for about six weeks, and then a slow abandonment that nobody formally decided on, just stopped happening one week and never resumed. The tool wasn't the failure. The maintenance was.
Personal knowledge management has existed as a recognizable hobby for years, with Roam, Obsidian, Notion, and Logseq each building devoted, vocal communities around elaborate note-taking systems, link graphs, and carefully maintained tags. Most people who tried one of these tools stopped, not because the idea was wrong, but because the upkeep cost more attention than most people were willing to spend on a system that mainly rewarded the effort of maintaining itself.
The Labs approach starts from a different premise: value should come from things a person was already doing — conversations with an assistant, drafts written for other purposes, decisions made and later revisited — rather than a separate practice of deliberate note-taking layered on top of an already busy life. ChatGPT's memory feature already does a smaller version of this, retaining details across conversations without asking a user to file anything away; the Labs project is testing whether that same passive-accumulation idea can scale into something closer to real infrastructure.
The people who already tried and abandoned Roam or Obsidian are the sharpest possible test group, precisely because they're the ones who tried the deliberate-curation version and found the maintenance cost too high the first time. If a system built as infrastructure rather than a hobby can surface something useful to them without asking for the curation that drove them away, that's a meaningfully different product than the last generation of tools.
The comparison also exposes what made the earlier generation fragile: those tools depended entirely on a user's continued motivation, and motivation is not a reliable foundation for something that's supposed to compound quietly over years without anyone tending it.
The open risk, still being tested internally, is whether passive accumulation produces something genuinely useful or just a large, unsorted archive that's technically complete and practically unusable — the same failure mode that thinned out the Roam and Obsidian communities, just automated instead of manual.
The hobbyist version of this idea asked people to become their own librarians, and most people, it turned out, never wanted the job. The infrastructure version is betting they just wanted their own past thinking to be there when it mattered, without having had to file it away themselves — and that bet either holds up under real use or it doesn't, which is exactly what the prototype is built to find out.