The Backlog of Possible Lives

For years I have been collecting notes and ideas that I thought I might return to later. Some were a few sentences, some were screenshots, links, names, half-written essays, domain names, rough product ideas, abandoned websites, old interviews, and things I noticed while traveling or working on something else. I kept saving them because they felt worth saving, but most of them required more time, skill, research, money, or concentration than I had at that moment. Writing them down became a way of postponing the decision about what to do with them. I could preserve the possibility without having to build anything yet.

The result is a very large backlog of possible things. It is spread across notes apps, old documents, browser bookmarks, cloud drives, messages to myself, and folders whose names made sense when I created them but no longer explain much. I do not really know how large it is. I only know that whenever I look through it, I find things I had completely forgotten and immediately want to start working on again. Some are obviously bad. Others are still interesting, but they became inaccessible because returning to them meant reconstructing the entire context around them. I had saved the idea but lost the mental state in which the idea made sense.

Digital tools became extremely good at helping us collect things and strangely bad at helping us return to them. It takes almost no effort to save an article, take a screenshot, record a voice memo, or write a sentence in a notes app. The act feels useful because the thought is no longer in danger of disappearing. But every saved fragment creates a small future obligation. At some point I am supposed to read it, understand it, connect it to something, decide whether it matters, and perhaps turn it into something. Those obligations accumulate much faster than the available time to resolve them.

After a while, the archive starts to feel slightly accusatory. Every unfinished idea represents a version of myself who expected a later version to have more time. There were things I wanted to build before I could code, things I wanted to write before I understood the subject, and projects that would have required a team when I was working alone. Some ideas arrived when I had neither the ability nor the confidence to test them properly. Others may have only seemed important because I was interested in them for three days. They all ended up stored in roughly the same way, which made it difficult to tell the difference.

AI is changing my relationship with this backlog more than it is changing my ability to produce completely new ideas. I already had enough ideas. The problem was that each one opened into several kinds of work I could not necessarily do. A note could require research, writing, visual design, software development, editing, translation, fact-checking, and publishing before it became anything another person could see. Even a small idea could quietly demand a week. It was easier to save it and move on.

Now I can return to a fragment and ask what it was trying to become. I can reconstruct the context, research the subject, compare it with similar ideas, find contradictions, test the language, and sometimes turn it into a working prototype before the initial interest disappears. Several notes written years apart can be placed next to each other and recognized as parts of the same question. An old interview can be transcribed and searched. A vague software idea can become something I can use instead of a paragraph describing something that might one day exist.

This feels less like acceleration and more like recovery. I am not only making new things faster. I am gaining access to work that was already partially done but had become unusable. The original noticing was there. The curiosity was there. Sometimes even the central argument was there. What was missing was the capacity to carry it through all the boring and difficult middle stages.

People are beginning to document versions of this. They are feeding years of diary entries into language models and discovering recurring themes that were impossible to see entry by entry. They are analysing thousands of old posts, organising chaotic personal knowledge systems, searching family archives, and turning large collections of notes into something they can finally navigate. The language around this is usually productivity, second brains, knowledge management, or AI-assisted writing. Those descriptions are accurate, but they make the change sound more administrative than it feels.

There is something stranger happening when a person can suddenly read their own past at a different scale. We have spent years creating huge personal archives without having a realistic way to understand them. The material was technically available, but practically lost. It sat behind search boxes that required us to already know what we were looking for. AI can begin with a much looser question. What have I repeatedly cared about? Which ideas kept returning? What did I notice before I had the language to explain it? Which projects were abandoned because they were weak, and which were abandoned because I was not ready?

I am interested in this because the backlog is not only unfinished work. It contains unfinished versions of a life. Every abandoned project points toward a direction that was possible at some moment. A film I did not make, a company I did not start, an article I did not finish, or a tool I could not build all belong to slightly different possible lives. Most of those lives should probably remain unrealised. There is no reason to believe every old ambition deserves resurrection. But until now, many of them were rejected by circumstance rather than judgment.

The cost of making things used to perform part of the selection for me. Ideas disappeared because I lacked time, money, technical skill, or access to the right people. That filter was crude, but it was effective. As AI lowers the cost of execution, I need a different way to decide what deserves attention. The fact that I can now turn an old note into a polished article does not mean the note contained enough thought. The fact that I can generate a working prototype does not mean the product should exist. AI makes it easier to confuse completion with value.

I already notice this when a fragment becomes too articulate too quickly. A few weak sentences can be expanded into something that looks finished and sounds more certain than I ever was. The result may be competent, but I do not necessarily feel that I discovered anything while making it. It can create a false history in which I appear to have had a developed argument when I only had an intuition. The backlog then becomes raw material for manufacturing a more coherent version of myself.

The useful part happens when AI helps me stay with the uncertainty rather than erase it. It can help me find the missing information, identify where an idea is generic, show me that somebody already explored it better, or connect it to another fragment I had forgotten. It can help me build enough of something to understand why it fails. That is different from asking it to complete everything. Sometimes the best outcome of returning to an old idea is finally being able to abandon it for a good reason.

This is why I have started to think of the backlog as an archive rather than a task list. A task list creates the expectation that everything should eventually be completed. An archive can contain dead ends, repetitions, unfinished thoughts, and things whose value only becomes visible much later. The goal is not to empty it. I doubt it can be emptied, especially now that making one thing usually creates five more possible directions. The goal is to make it possible to move through it without being crushed by it.

AI has not reduced the number of things I want to do. It has probably made the problem worse because many projects that once seemed unrealistic now look temporarily possible. But the backlog no longer feels like a sealed storage room. I can enter it, ask questions, combine fragments, test old assumptions, and occasionally bring something back into the present.

The interesting question is therefore not what AI will help me invent next. I already spend enough time thinking about that. The more useful question may be what I have already been trying to make for years, and whether I can finally understand why I never made it.

Connected reading

• My Screenshots Know About the Life I Almost Lived
• Now I’m Limited by What I Can Imagine
• Raw Drafts Are a Form of Learning
• The Prototype Has Become Too Cheap to Trust
• The Real Bottleneck Is Human Attention

Sources

• Digital hoarding as a new form of obsessive-compulsive disorder?
• I Kept a Diary for Seven Years. An LLM Finally Read It.
• My Second Brain: Markdown, Dropbox and AI
• The Value of Creative Clutter
• AI writing support and psychological ownership

AI authorship note

This essay began with Kris Haamer noticing that AI was helping him move through a long backlog of notes and ideas he had lacked the time, skill, or capacity to build. AI helped research how other people are using language models to revisit diaries, old posts, personal knowledge systems, and creative archives, then supported drafting and publication. Kris repeatedly pushed the piece away from short, polished idea units and toward longer connected paragraphs in his own exploratory voice.