I built Obsidian in Slack before Obsidian existed. Since 2016, I have been using Slack as a personal knowledge base -- not for work teams, for myself. Separate channels for strategy, AI, design, food, retail, Web3, books, jobs, trends: whatever I was thinking about, the links went there. One message, one URL, no annotations needed. The act of saving was the point.

Last week I extracted the full archive and ran it through a categorisation script. 33,643 URLs. 474 channels. A decade of intellectual residue, made searchable.

Here is what it showed.

The shape of a decade

The volume curve tells its own story. In 2016 I saved 223 URLs. By 2019 that had grown to 2,558. By 2021 it peaked at 6,559 -- nearly double the previous year, and the single highest volume in the archive. Then it fell. Not because the habit ended, but because the nature of what I was reading changed.

The peak is not random. 2021 was the Web3 year. Crypto/Web3 content went from under 2% of my reading in 2020 to 17% in 2021. Every NFT project, every think piece, every long-form argument about ownership and provenance -- all of it is in there. The channel was #brands-nft. It had 1,878 links. It is a complete record of an obsession.

By 2024, Web3 was back below 5%. By 2025, 1.5%. The market moved. The attention moved with it.

Going deeper, not wider

The more interesting signal is not volume. It is range.

In 2021, at peak volume, I was reading from 1,816 distinct domains in a single year. By 2025 that had dropped to 641. Less than half. The number of URLs saved is comparable year-on-year -- the difference is that the sources have concentrated.

That is not a decline in curiosity. It is a change in how curiosity operates. The wide-net phase gives way to the deep-well phase. Fewer sources, better questions, more time with the things that actually compound.

The archive has a circadian rhythm

Because every link was saved with a Unix timestamp, I can see exactly when the thinking happened. The pattern is unmistakable.

Wednesday is the biggest day -- 5,701 links across all years, 28% more than Sunday. The week builds to a mid-week peak and never fully stops at the weekend (Saturday: 4,095, Sunday: 3,962).

By hour, there are two distinct peaks. A morning surge between 9am and 11am. A quiet midday dip. Then an evening surge that builds from 9pm and peaks at 11pm. Only 25 links were ever saved at 4am. The archive has a genuine sleeping window.

April is the busiest month -- 4,264 links, nearly double January's 2,362. And it is not spread evenly across the month. Week 15 of the year (mid-April), across all years combined, is the single busiest week in the archive: 1,195 links. Spring is when the thinking arrives.

One in four links is a tweet

X and Twitter combined account for 8,622 URLs across the decade -- more than one in four of everything saved. LinkedIn adds another 3,988. Strip those out and the actual reading diet becomes visible.

Substack combined across publications: 811. Medium: 519. YouTube: 560. Then individual titles: The Verge (258), TechCrunch (234), Bloomberg (172), The Drum (173), Adweek (144), BBC (121), Hypebeast (112), Fast Company (107).

The Drum and Campaign Live sitting alongside Bloomberg is a signature of the agency years. Hypebeast and Vogue Business (89) sitting alongside TechCrunch tells you what the brief actually was.

The turn

The data point I did not expect: mikelitman.me appears 359 times in 2026 alone. In every previous year combined, it appears 5 times.

For nine years, the archive is a record of reading other people's work. In 2026, it starts recording my own. The fifth most-linked domain in the entire archive, almost entirely from a single year. That is a direction of travel written into the timestamps: from observer to maker.

What links reveal that content does not

Content is edited. Links are honest. When you publish something, you choose what to show. When you save something, you are recording an unfiltered response to an idea -- the moment before the edit, before the opinion hardens.

33,643 of those moments, across ten years, add up to something a CV cannot capture. The intellectual biography is in the URLs. The obsessions are in the channels. The working rhythm is in the timestamps.

I built the full interactive version of this archive at mikelitman.me/decade-in-links. Every number on that page comes from the raw data. No estimates.