Addictive Agents

Everybody loves Claude Code. It’s so good that almost anybody is now empowered to build an app. Seemingly every conversation I have with other tech folks is about what they’re building and how easy it has been. The excitement is palpable.

The results seem very positive: families are building apps to track chores and reduce internal friction; fund managers are creating dashboards to track a crisis; and some of our portfolio companies are producing years of work in a matter of weeks. At Sterling Road, numerous internal apps, built by different team members, are already live.

In fact, we’re making so much progress that I start on the various projects right after I wake up. I would keep working on them until I went to bed, but my wife insists I stop at 8.30pm. I’m not alone, many people I spoke to in the last 2 weeks are also building day and night.

We see similar behavior on the consumption side with Character.AI, where some customers spend 12+ hours a day with their Agent. Of course, it’s much easier to accept compulsive behavior when it’s productive and you control the AI’s behavior. But if AI Agents are already this addictive, how much longer will we be in control?