Feels like nVidia is trying to get their name in the OpenClaw game, hard (nemoclaw, openshell) but why? So they can sell local inference hardware?

https://www.nvidia.com/en-us/ai/nemoclaw/

Seeing nVidia go so hard into OpenClaw ("Every company is going to need an OpenClaw strategy") is bizarre. LLMs, I get the value. Same with agents. But OpenClaw is basically an agent with computer use and without most of the guardrails.

That's what makes it powerful, but also makes it dangerous. nVidia is trying to put the claw in a box with OpenShell and gateways, but by the time you get it locked down to a safe level, isn't it just another agent like Claude Code?

My little Linode Kubernetes cluster is such a great way to deploy the little web apps I keep coming up with. Today's is a fairly simple diet tracker that matches what I want. And figured I'd try a Passkeys-only login system.

I have a $24 4gb node that's running 6 Django apps, Postgres, the ingress, certbot, and a few other things.

https://diet.fallday.ca

(Yeah the Y axis is redacted)

Claude's /loop is worth remembering. It's not a direct command so much as direction to Claude that it should set up a periodic job.

I made a DNS change and the next step depends on it propagating, so..

/loop 1m check for dns propagation

(Assuming it knows the domain from the context) It will just wake up every minute, check, and go back to sleep, or delete the timer and continue the task that was blocked.

Freedom Mobile has been a fun ride. I started at a much higher rate and every time there's a promotion I jump on it so my price has been inching down since I signed up; now I'm at $40/month for 100GB of data in Canada and 10GB of global roaming.

Coverage has never been a problem for me but I live in the GTA so I guess that's no surprise.