Nov 12, 2026
Introducing Current
Introducing Current, an opinionated way to work with your agents.
Current lets you run hundreds of agents while staying in control. I've built everything with it over the past couple of months, including Current itself.
You are the bottleneck
The problem with running agents in parallel isn't their intelligence anymore. It's you, and your attention. I kept scrolling walls of text and tool calls, looking for answers I'd already been given, and losing track of what was done.
Outputs, not inputs
Current focuses on your agents' outputs, not their inputs, by dropping the chat entirely. Everything is an issue, from a quick fix to a big project with sub-issues. You write it, and your agents pick it up and deliver.
When your agents are done, they notify you and show you a deliverable: a report, a diagram, a chart, a before and after video, or an interactive app that shows their work. You review the work, not how they got there. If you want a change, you comment, and the agent goes back to work.
How it changed the way I work
Current completely changed how I work. It made me rethink what I can build and become much more ambitious. Before it I made about 40 to 50 GitHub contributions a day. This week I've been at 500+ a day.
It was never meant to be a product
Current started as a skill file begging Claude to understand I wasn't reading its messages anymore, and that whatever it did, it had to prove the work was delivered. That file kept growing. I put a UI on top of it, realized I spent my whole day in that index.html on my desktop, and took a few days to rebuild it properly.
Try it
Current is live and free on Mac, Windows, Linux and iPhone. It works with Claude Code, Codex, OpenCode and many more agents. Teams, for shared workspaces, is $99 a seat a month or $89 billed yearly.
There's a lot more coming. Multiplayer, the iPhone app and more each get their own post on the changelog.