The work graph is the next database

Every tool in your stack records its own artifacts. Git stores commits. Jira stores tickets. Confluence stores docs. Slack stores conversations. Each one is a well-structured silo of what was produced.

But nobody records the activity between them: what you actually did, in what order, and why. The commit doesn't know about the Slack thread that prompted it. The ticket doesn't know which three files you opened before you understood the bug. The doc doesn't know it was written after a 40-minute debugging session that changed your mental model.

The missing layer

I've been building systems that capture this — a temporal graph of real work. Not metadata about artifacts, but the actual sequence of human and machine activity that produced them. File opens, terminal commands, browser tabs, agent conversations, git operations — all timestamped, all connected.

This is what I'm building with WorkLens: a developer-first watcher that captures the rhythm of work as it happens. Not for surveillance — for context reconstruction.

Why this matters for AI

When an AI agent joins your team, the first thing it asks (implicitly) is: "What's been happening?" Today, the best answer we can give is a pile of artifacts — READMEs, recent commits, open tickets. That's like handing a new hire a filing cabinet and saying "figure it out."

A work graph answers the real questions:

• What was the team working on this week — not just what shipped, but what was attempted?
• Why did we choose this approach over the alternative?
• Who has context on this subsystem, and when did they last touch it?
• What's the actual cost (in time, in context switches) of this kind of change?

This is the substrate AI teammates will query. Not your codebase — your work history.

The thesis

Every major platform shift creates a new foundational data layer. Relational databases organized business transactions. Document stores organized web content. Graph databases organized relationships. The next one organizes work itself — the temporal, multi-modal, cross-tool graph of what a team actually did.

That's the thesis I keep pulling on. The model matters less than the data it can reason over — and right now, the richest data about how software gets built is evaporating the moment you close your laptop.