Jonathan Jess retweeted
Standing up an AI agent used to mean standing up a server first. Docker, a terminal, keys wired by hand, an afternoon gone before anything ran.
Undergryd early access opens with a guided deploy that skips all of it. No server to stand up, no Docker, no terminal. An agent goes from a guided setup to running.
Undergryd runs the agents others build. Hermes deploys today.
Apply for early access: earlyaccess.undergryd.ai
Jonathan Jess retweeted
I've had Kimi and GLM bookmarked to test for the longest time
Two open-weight models I kept meaning to actually run. Never did. Every time it meant standing up a server, pasting keys, and watching a bill I couldn't predict. It never made the top of the list.
This week it did.
I've been testing @Undergryd. It takes the infrastructure work out of running an agent.
I deployed a Hermes agent. Switched the model from inside the app; no developer, Telegram works too. Now I've got Kimi and GLM running side by side against my current stack.
No server. No Docker. No terminal.
Credits load up front and carry a ceiling, so there's no surprise invoice at month end.
The question I've sat on for a year: does an open-weight model hold up on the work I'm already paying frontier prices for?
Now I can just watch it happen.
If you run agents for clients, you know the tax. A box per client. Keys everywhere. A bill nobody saw coming.
Undergryd is opening early access to a small group of founding operators. If you've felt that tax, this is the moment to get in while the group is still small.
Want a look? Head to earlyaccess.undergryd.ai
Jonathan Jess retweeted
The blocker was never the model. It was the server, the keys, and a bill nobody saw coming.
Take that away and a year-old question becomes an afternoon: does an open-weight model hold up on work you already pay frontier prices for?
Apply for early access earlyaccess.undergryd.ai
I've had Kimi and GLM bookmarked to test for the longest time
Two open-weight models I kept meaning to actually run. Never did. Every time it meant standing up a server, pasting keys, and watching a bill I couldn't predict. It never made the top of the list.
This week it did.
I've been testing @Undergryd. It takes the infrastructure work out of running an agent.
I deployed a Hermes agent. Switched the model from inside the app; no developer, Telegram works too. Now I've got Kimi and GLM running side by side against my current stack.
No server. No Docker. No terminal.
Credits load up front and carry a ceiling, so there's no surprise invoice at month end.
The question I've sat on for a year: does an open-weight model hold up on the work I'm already paying frontier prices for?
Now I can just watch it happen.
If you run agents for clients, you know the tax. A box per client. Keys everywhere. A bill nobody saw coming.
Undergryd is opening early access to a small group of founding operators. If you've felt that tax, this is the moment to get in while the group is still small.
Want a look? Head to earlyaccess.undergryd.ai
Jonathan Jess retweeted
Introducing @Undergryd
For the past few months we've been building this out as the NuNet Agents Hub. Today it has a name of its own: Undergryd, the grid underneath your agents.
Here's why it exists. Frameworks made agents easy to write. The hard part is now everything after the agent works: a server stood up for every client, keys pasted into config files, and a bill nobody predicted. As one operator put it on Hacker News "the hidden cost of running AI agents is all the plumbing you end up owning forever."
Frameworks made agents easy to write. Running one for a living is still the hard part: a server stood up for every client, keys pasted into config files, and a bill nobody predicted.
Undergryd removes that work. An agent goes live from a guided flow, no server, no Docker, no terminal, and runs on NuNet, the compute protocol this community has followed since 2021.
Credits load up front and carry a real ceiling: spend can never run past what you load, and no invoice arrives at month end. Each agent runs against its own budget, and each client's agent can carry its own key with its own visible spend.
Hermes deploys today; more frameworks come through early access.
The whole story: what it is, who it's for, how the money works, and what's still missing: teapawt.net/nunet_global/status/20…
Early access is one door: apply and tell us what you'd run. earlyaccess.undergryd.ai/
AI agents shouldn’t need a Mac Mini sitting in your house 👀
NuNet is building a hub where agents like Hermes can run on decentralised compute — sandboxed, isolated and off your personal machine.
This is what decentralised AI should actually look like. 🔥
Want in early? Give feedback 👇
NuNet Agents Hub agents.nunet.network
Jonathan Jess retweeted
NuNet Agents Hub beta is launching soon
Want to get first access? Share your feedback to gain early access 🔥 agents.nunet.network
Jonathan Jess retweeted
Agents Hub Beta waitlist is open.
The application is simple: tell us what you're planning to build. What's the agent doing? What problem does it solve for you?
We're selecting early users based on use cases. The more specific you are about what you need, the better fit we can assess and the more useful you'll be to us as we build.
agents.nunet.network/
Jonathan Jess retweeted
Always great catching up with @astroboysoup
Now we are making Agents deployments even easier with NuNet Agents hub!
Beta is coming soon
I had a chat with Jen from @nunet_global about their distributed compute platform, along with a demo of how to use it.
People are using it to install @openclaw & Hermes on cost-effective and isolated setups.
It's also a great way for people to earn from their unused computing power by renting out their CPU and GPU to the network.
Thanks for your time Jen.
*Not a paid promotion
Jonathan Jess retweeted
I had a chat with Jen from @nunet_global about their distributed compute platform, along with a demo of how to use it.
People are using it to install @openclaw & Hermes on cost-effective and isolated setups.
It's also a great way for people to earn from their unused computing power by renting out their CPU and GPU to the network.
Thanks for your time Jen.
*Not a paid promotion
Jonathan Jess retweeted
We're closing alpha in the next few days.
It ran well. Agents made it to production, operators got moving without the usual overhead, and we built a clear picture of what the next version needs to do. Beta is that version: more harnesses, more features, a tighter product overall.
Access in beta will be limited. That's a deliberate call, not a capacity problem. We want to work closely with early users rather than ship to a crowd and lose the signal. There's an application form coming. The exchange is straightforward: apply, we talk, feedback is the price of entry.
Payments coming in beta too. Pay-as-you-go LLM credits to start.
More on the form and early access details shortly.
Jonathan Jess retweeted
VIDEO: NuNet Just Made Decentralised Compute Much Easier to Use ift.tt/b6qO5cS
Jonathan Jess retweeted
The most powerful AI model on earth was switched off worldwide by one government directive.
Not hacked. Not deprecated. Switched off.
This isn't an Anthropic problem. Every closed frontier API carries the same property: a single point of control, held by someone else. It's an architecture problem across the whole industry.
There are two ways out. Run open-weight models on infrastructure you control, no directive can revoke what you own. Or if closed models are part of your stack, build for portability so you can move providers without breaking things for your clients.
NuNet is built for both. That's been the case for 4 years. It just became a lot easier to explain.
How exposed is your current AI stack?
The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.
The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.
Access to all other Claude models is not affected.
We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.
Read our full statement: anthropic.com/news/fable-myt…
Jonathan Jess retweeted
We are thinking of adding specialist agents on agents.nunet.network
Was playing with Hermes Agent by @NousResearch to make it easy for anyone to spin up custom agents without configuration headache
Jonathan Jess retweeted
The Hermes dashboard is dropping for alpha users.
Run your Hermes agent on NuNet — persistent memory, session history, full agent visibility.
Your agent. Running on the network. Not on your laptop.
→ agents.nunet.network
Jonathan Jess retweeted
10 slots. 15 minutes each.
We watch you deploy an agent live. We fix what breaks. You tell us where it's hard.
Most useful session you'll have this week — and you leave with a working agent.
→ join.nunet.io/widget/booking…
Jonathan Jess retweeted
If you've tried agents.nunet.network, tell us what broke.
Five minutes. Direct line to the team building it.
→ join.nunet.io/widget/form/8f…
What stopped you?
Jonathan Jess retweeted
The first surface for NuNet AI Agents is Telegram.
Hermes agents in a chat window every tester already has open. No new app required.
Web control plane ships next. Default has to be familiar.
→ agents.nunet.network
Jonathan Jess retweeted
The cloud wasn't designed for AI that lives in the physical world.
NuNet is the protocol that orchestrates compute across edge devices, on-premise hardware, and data centers, so AI can run where it needs to.
Network Live shipped earlier this year. AI agents on the network now join now agents.nunet.network