Run NVIDIA’s latest 120B model on Lightning AI
NVIDIA just released Nemotron 3 Super, a 120B total parameter model with only 12B active parameters, built on a hybrid Mamba-Transformer MoE architecture.
This is a big deal for devs building agentic systems because the sparse activation means you get strong reasoning at a fraction of the compute cost of a dense 120B model.
Lightning AI is a day-zero launch partner, and they’re already serving the model at ~484 tokens/s, making them one of the fastest inference providers for it.
The model also supports a 1M token context window natively, which means your agents can load entire codebases or long document sets without chunking hacks.
If you’re looking to try it out, Lightning AI has a hosted endpoint ready to go →
Thanks to Lightning AI for partnering today!
Set Up a Secure OpenClaw Deployment!
If you’re deploying an AI agent on a VPS, the default setup is not secure.
Default SSH config exposes root, all ports are publicly accessible, and there’s no network isolation. For an agent that can execute code and interact with external tools, that’s a real risk.
We recorded a video covering how to deploy OpenClaw on a DigitalOcean VPS and lock it down end-to-end.
Here’s what we’ll cover:
DigitalOcean Droplet setup + getting OpenClaw running
Connecting your agent to Telegram
Tailscale: making your server invisible to the internet
Locking down SSH + creating a non-root user
Firewall: blocking everything except Tailscale
Set up a secure tunnel: Your machine → VPS
Execution policies: going from chatbot to full agent
Adding custom skills
Use cases and going from 1 to 10 agents
The security fundamentals that you’re going to learn here generically apply to any personal AI assistant or VPS setup.
We use DigitalOcean because it gives us authenticated gateway tokens, firewall-level rate limiting, non-root execution, container sandboxing, and private DM pairing out of the box.
With Tailscale and DigitalOcean’s firewall configured together, the server has zero public-facing ports. The only way to reach it is through your private Tailscale network.
You can create your OpenClaw Droplet on Digital Ocean here →
Thanks for watching!
