Most weeks I end Friday thinking I have not done much. Then Sunday evening the digest arrives. Five blog posts. Five decks. A dozen site updates. I did all of that. I just forgot.
Since 2016 I have kept a personal Slack workspace called Mike's Links. Just me, 459 channels, and a habit of saving links into named folders as I move through the internet. Not a team tool. Not a project manager. A filing system for where my mind goes: #ai-artificial-intelligence, #cultural-interface, #jobs. I never set out to build anything systematic. I just started filing, and never stopped.
Last month I ran a Python script across ten years of that data. 3,312 messages in a 90-day window. A visible AI inflection in 2022-23. A pattern of annual behaviour that told a story about my career I could not have written from memory. I turned it into a dashboard. Then I built a weekly digest. The dashboard was interesting. The digest was useful.
The digest that reads your mind
Every Sunday evening, a script runs automatically. It scans every channel in my workspace for the previous seven days, counts messages, extracts URLs, and generates a broadsheet-style report at mikelitman.me/week. Top channels, top domains, cross-channel patterns, a Claude-generated editorial line that notices something honest about the data. The whole thing takes about ten minutes to run and lands on Sunday evening without me touching it.
What I did not expect was how clearly it would show me what I had been tunnel-visioned on. When I am deep in a build, I think that is all that happened that week. The digest arrives and it turns out I also saved 40 things to #cultural-interface, drafted notes in #ai-artificial-intelligence, and apparently had an entire parallel intellectual life running alongside the shipping. The week does not feel like it contained all of that. The data says otherwise.
The week of 13-19 April, the digest showed 256 saves across 19 active channels. I had been convinced the whole week was one thing: preparing and running a live AI demonstration to 125 people. The digest showed a different week entirely. Richer, more scattered, more alive than the version I had been carrying around.
But it showed me something else too. It was not just what I saved. It was what mode I was in.
Three modes, clearly legible
After a few weeks of digests, a pattern became impossible to ignore. The channels that dominate in any given week are not random. They cluster around three distinct states, and each state has a clear signature in the data.
When #cultural-interface and #ai-artificial-intelligence lead the week, I am in consuming mode. Lots of reading, lots of saving external material, high volume of links from publications and X. This is the mode where ideas accumulate. Where the raw material comes in. Usually it precedes something being built.
When #claudecode-agents, #buggysmart, #ml-the-system, and project-specific channels dominate, I am in building mode. The consumption drops sharply. The saves are mostly reference material for active work: documentation, examples, things to revisit while shipping. That week in April was building mode. I had just run a live AI demonstration to 125 people and was deep in following up.
When #jobs, #cv, and #to-pitch spike, the mental state is different again. Not exactly anxiety, but proximity to it. Income pressure, career questions, the specific flavour of attention you give to your professional situation when it is not quite settled. The workspace has never lied about this. The data is honest even when I am not being honest with myself.
What makes this genuinely interesting is the gap between the calendar and the channels. The calendar shows what I planned to do. The channels show what I was actually drawn to. Sometimes they match. Often, they do not. The week I thought I was focused on strategy, the channels showed I was deep in #jobs. The week I thought I was taking it easy, #claudecode-agents had 61 entries. The saves do not reflect intention. They reflect preoccupation.
Nobody else is using it this way
I have shown this system to a lot of people over the past year. The universal reaction is some version of: I had no idea you could do this. Most people use Slack as a communications tool. Messages in, messages out. Notifications, threads, integrations. A faster email with worse search.
The thing I built is structurally different. The channels are not conversations. They are named attention folders. The act of posting a link to #cultural-interface is not communication with anyone. It is a filing decision. It says: this belongs in the collection of things I am paying attention to under the heading of culture. Over time, the filings add up to a map. Not a map of what I know, but a map of where my mind has been.
Spotify Wrapped tells you what you listened to. This tells you what you were thinking about. The distinction matters. Music is something that happens to you. A save is a choice. Every link in every channel is a moment where I stopped, decided this was worth keeping, and filed it under a name I gave it myself. The taxonomy is mine. The pattern it reveals is mine. No algorithm decided what goes where.
The most honest record I keep
Across a quarter, the digest becomes something more than a weekly reminder. Mode cycles appear. The building sprints, the consuming phases, the anxiety windows. You start to see your own rhythm. And underneath all of it, the consistent gap between how productive you feel and how productive you actually were.
Time-tracking tools tell you where the hours went. This tells you where the mind went. And where the mind goes, before you have done anything deliberate about it, is where the work eventually comes from.
Ten years of links. A decade of filing. And the most useful thing it has produced is a Sunday evening report that tells me: you did more than you think. You went more places than you remember. The version of your week that lives in your head is always edited, always flatter than it was. The channels are not.
I set out to build a knowledge system. What I built was the most honest record I keep of myself.
