At the end of last November, Clawdbot launched and within weeks became one of the fastest-growing open-source projects in GitHub history. Everyone around me was installing it; I installed one too and let it take over my chat gateway. Three days later, I uninstalled it.

Today it's called OpenClaw. The third name so far.

1. Two renames in three months

The timeline goes roughly like this: late November 2025, Peter Steinberger ships Clawdbot, the name a pun on Anthropic's Claude, mascot a lobster. Late this January, Anthropic's lawyers send the notice — the name is too close to Claude — so it becomes Moltbot. Three days later, presumably after realizing Moltbot sounds like a washing-machine model, it becomes OpenClaw.

The most surreal interlude happened in the rename gap: in the seconds between releasing the old handle and registering the new one, squatters moved in, and the fake accounts pushed a CLAWD coin to a peak market cap of sixteen million dollars before it went to zero. That same week, the project's Discord was busy kicking out anyone who mentioned crypto — allegedly even typing "bitcoin" got you banned.

A project that can't hold onto its own name for a week clearly has no "think it through" slot on the calendar.

2. The slop isn't in the code, it's in the structure

The security community's scorecard: back when it was still Clawdbot, researchers counted 512 vulnerabilities, 8 of them critical; Bitsight found over a hundred and thirty thousand instances exposed on the public internet with credentials lying around in cleartext; the ClawHub skill marketplace yielded 386 malicious "skills," uniformly teaching you to trade crypto. Plus that one runaway incident that made the front page of WIRED.

Some say this is the normal growing pain of open source. I disagree. Behind all these numbers is the same structural problem: this is a gateway that hangs on your accounts and credentials 24/7, and it grew up "run first, ask later."

Piling up code volume with AI is easy; thinking through trust boundaries is hard. That's the very definition of slop — the output capacity went up, the judgment didn't.

3. Why it's doomed, specifically

Not because it's bad — because the foundation of this model is rented:

  • The bulk of its value is the underlying model, not this shell. Every model upgrade shaves the shell's value down another notch.
  • The moment an official assistant ships "connect to your chat apps" as a product feature — and that day isn't far — every wrapper's reason to exist gets vacuumed out at once.
  • Hype-driven open source has no maintenance will. Moving into a foundation is damage control, not a renaissance.

To be clear: I'm talking about the pattern, not the person. Steinberger is a good engineer who just picked the steepest route on the course and then drove it at full speed.

4. Closing

Things that last are built on boring engineering and restraint: think through boundaries, write tests, learn to say no. Slop has exactly one weapon — speed.

The direction of the tide is already clear. The moment the models get a little stronger, the first things washed away will be these shells.