Read my rant about Antropic and Claude models here:

https://paste.orangefox.tech/69d0b640202a139f3e4b3181

I tried to cover my view over topics such as:
- Department of Defense, and why Antropic is not as good as press covers it;
- Developer-relations and constant ban of users;
- Open-weight models contributions of Antropics (none of it);
- Why people think Claude models are more advanced.

If anyone wonders, the post was written by me without using AI tools other than proofreading.

So I played with openClaw yes, even though it was buggy (clearly it was vibe coded), I found that it's pretty nice to do the agentic work like gitlab issues management, email, and other "assistant"-related workflows.

Once I tried to delegate coding tasks, it failed so miserably, I felt like it was just a joke.

It made a GitLab MR, tried to fix CI/CD by pushing changes to unrelated branches, many times, and at the end it lost all the work.

I will continue playing with it as an assistant, but I don't think it can replace a programmer yet, at least one that isn't an intern.