Lines of code. Commits. PRs. Tickets closed.
None of these signals work anymore.
We're working with coding agents now. There are new signals around them — and the old metrics can't see any of it.
So I started building toqn to answer one question: how many prompts did I send to my coding agent today?
I had a hunch this was one of the strongest signals of my productivity.
I didn't know whether sending more prompts means you're more productive, or fewer. But I knew the simplest new signal wasn't how many tokens I spent in a day. It was how many prompts I sent.
So I built a simple hook.
After a couple of days, I realized it was a genuinely good signal. On the days I felt like I'd shipped a lot, there was a clear positive correlation between prompts sent and work done.
Then I started using skills that force the agent to keep going on its own. And I was back to square one. I was doing more with fewer prompts. The correlation broke.
So the next step was natural. Look at everything else.
Token count was always going to matter. But never on its own.
I decided to build a dashboard that watches every move my agent makes. Not the text inside the prompts. Not the code it generates. Just the signals around it.
Tools called. Files read. Files changed. Lines added and removed. Bash commands, grouped by what they do. Git commits. Git pushes. PRs opened. Subagents spawned. Skills triggered. Tool errors. Tokens in, tokens out. Cache hits. All of it, together.
Then I went further. I added generated suggestions, insights, warnings.
Now I can even answer which agent and which skills make me most effective on a given project.
The result: a much better picture of how I actually worked in a given day.
The next generation of productivity signals lives in your agent's actions. And in how, and how much, you interact with it.
toqn.dev — the signals that actually matter.