Most people should not use OpenClaw (yet)

Every published a guide to OpenClaw called Claw School. The vibe throughout is "it just works". That hasn't been my experience at all—I've run into a lot of bugs and faff, and others report the same. If I didn't have a background in software dev, it'd have been pretty unusable.

Even given that background, ex post it'd have been better to wait a few weeks, rather than starting in late-January. The project is advancing quickly, it's far more stable than before, but still... very early.

Hosted Claw services—that promise to make everything easy—are popping up. I tried two—MyClaw and Klaus—and immediately hit blocking bugs that couldn't be resolved without a support ticket. They're not ready yet.

So:

If you've some software engineering experience and you're thinking of giving OpenClaw a try this month, my take is "reasonable, but not clearly above bar". Probably, you're better off investing in your agent skills and personal software tools, and reconsidering the case for a Claw in mid-April.
If you don't have SWE experience, skip it for now. Consider again in mid-April.

A good hosted Claw service—or a better product from a mainstream AI company—will likely show up in a month or so.

Appendix 1. How am I using OpenClaw?

I use it a couple of times per day. Main sources of value:

Do stuff while AFK. Kick off simple coding tasks or make plans from my phone. This affordance shifts more tasks into the "think of it => do it immediately" category. Screenshotting error alerts and saying "please fix" is nice.
Mochi flashcard review. Five cards are delivered to my phone each morning. Interactive review in chat—with the ability to discuss cards with the model—is very nice.
Morning briefing. My chief-of-staff skill writes a morning brief. It's nice to read it over breakfast, before I get to my desk.

Health warning: AI on your phone is very addictive, and OpenClaw is especially bad since it messages me proactively. I block my AI phone apps from 9pm - 8am, and all day Sunday.