I have Raindrop bookmarks tagged #vip, #now, and #someday. I have Trello boards with lists called “This Week,” “Next Up,” and “Later.” I have an Obsidian vault with folders whose purpose I can no longer remember. I have starred Claude chats I will never reopen. I have a browser with 47 tabs across three windows, each one a promise I made to myself that I would “come back to this.”

None of it is working. All of it feels like it’s working. That’s the trap.

I’ve been thinking about why, and the answer turns out to be uncomfortable: the problem isn’t that I don’t have a good system. The problem is that every system I build is actually a system for avoiding decisions.

Saving something takes two seconds. Deciding whether it matters takes thirty. Deciding you were wrong to save it, and deleting it, takes something harder to measure, because it requires admitting that the past version of you who hit the bookmark button was optimistic, or bored, or avoiding something else. Deletion isn’t about the link. It’s about the person who saved it.

So we don’t delete. We demote. #vip becomes #now becomes #someday becomes a folder called _archive that we never open. And because demotion is easier than deletion, everything drifts downward, and the tiers stop meaning anything. Within a year, #someday is just a graveyard of dead enthusiasms, and #vip is everything you haven’t gotten around to saying no to yet.

This is what I’ve started calling priority inflation. Every label that promises to mark the important stuff ends up marking everything, because the act of saving already felt like a commitment, and downgrading a commitment feels like a moral failure. So the labels degrade until they carry no signal at all, and you add a new one, #actually_now, and the cycle restarts one layer down.

The failure isn’t in the tools. You could burn Raindrop to the ground and switch to Pocket, Instapaper, Obsidian, Notion, pen and paper. You’d reproduce the exact same structure within six months, because the structure is downstream of a psychological need, not a tooling gap.

The need is this: we want our attention to count for something. When I bookmark an article, I’m not really saying “I’ll read this later.” I’m saying “this resonated, and I want there to be a record that it did.” When I star a Claude chat after a weekend of ideation, I’m not planning to come back to it. I’m asking for acknowledgement that the work happened.

And any system that can’t give you that acknowledgement without demanding you actually follow through will eventually drown in your own guilt.

Two different jobs

The move that actually helped me was realising those are two different jobs.

Things I might need again is a retrieval problem. It’s solved by search. Modern search is very good. If I can remember three words from an article I read in 2023, I can find it in ten seconds. I do not need to curate it. Curating it is theatre. It makes me feel like I’m being responsible about information, while actually just adding noise to a pile I’ll never revisit.

Things that mattered to me is a completely different problem, and it has nothing to do with retrieval. It has to do with respect. The starred Claude chat doesn’t need to be findable; it needs to exist somewhere that acknowledges it was real. A named folder, a project page with a title, a line in a list: the container is the whole point. You might never open it again. That’s fine. The naming was the act of respect.

Once you separate those two jobs, the whole thing gets simpler.

For retrieval: capture aggressively, triage daily, delete without ceremony. Most of what you save will be deleted within a week, and that is the correct outcome. The small amount that survives isn’t surviving because it’s important. It’s surviving because you used it recently. Recency is a much better signal than you’d think.

For respect: name things. Give each real interest, each genuine project, each weekend-long obsession its own container with its own name. Let it rest. Don’t archive it, don’t demote it, don’t promise yourself you’ll come back. Just let it exist, named, in a place where you’ll occasionally walk past it and nod. That’s enough.

The cost of the ninety percent

The hard thing is accepting that “enough” is enough. Our tools keep trying to sell us on compounding. Every system is better next year, every note connects to every other note, nothing is ever wasted. It’s a seductive promise and it’s mostly a lie. Ninety percent of what I’ve ever saved, I’ve never used. Ten percent I’ve used once. A tiny fraction has compounded into anything. And that’s fine. The ten percent is worth the cost of the ninety.

What I keep coming back to is that the healthiest version of a personal knowledge system is one that forgives you for not using most of it.

Most bookmarks will die. Most notes will go unread. Most projects will go dormant, and most dormant projects will never wake up. This isn’t a failure state. It’s the normal shape of a curious life. You follow threads, some of them go somewhere, most of them don’t, and the ones that don’t are still part of how you got to the ones that did.

A system that makes you feel guilty for this is worse than no system at all. A system that lets you capture, name, and let go, one that gives the past version of you a respectful send-off instead of an accusatory to-do list, is one you’ll actually live with for years.

What it’s actually about

I’m still figuring out the mechanics. Folders, tags, frontmatter, cron jobs, the usual tinkering. But the mechanics were never the interesting part. The interesting part is learning that the bookmark tagged #vip from eighteen months ago isn’t a broken promise. It’s just a memory of caring about something, once. It deserves to be remembered. It doesn’t deserve to follow me around forever.

You can delete it. You can move it to _attic. You can let it rest in a folder called dormant without apology. The past version of you who saved it was doing their best with what they knew. You’re allowed to honour that and move on.

That’s the whole trick, really. Honour and move on. Most of what fails in personal knowledge management fails because we can do the first part but not the second.