Agent OS: Fable 5 vs Fusion, Windows Setup, and Free APIs artwork

Agent OS: Fable 5 vs Fusion, Windows Setup, and Free APIs

AI News Today | Julian Goldie Podcast

July 3, 2026

Agent OS Q&A: Fable 5 Tests, Hermes Setup, VPS/Windows Tips, and Free API OptionsThis episode answers community questions about an agent operating system setup with memory and multiple agents (including newly added NotebookLM short videos), covering what people are using Fable 5 for, its...
Speakers: Julian Goldie
**Julian Goldie** (0:00)
Today, we're going to be answering the latest questions on our agent operating system, how to build them, what to use, how to set up Hermes, et cetera. And I'm going to run you through step by step everything that we've learned so far based on the questions from the community today. So you can see this is an agent operating system where we have like our memory plugged in, we have all our agents working together, we can add new features, for example, like NotebookLM just dropped short videos yesterday, we already plugged that into the system. We have all sorts of cool stuff, even a music agent and a video agent, and everything else plugged into this system. So let me get straight into some of the latest questions we've had, and see what we've got here so far. So one of the first ones we got actually was what do people use in Fable 5 for? So one of the first things I'd say here is like Fable 5 pretty limited. I saw like some people posting that they used about two prompts and then it ran out of credits. So I think the token usage on it is pretty limited depending on what plan you're on. So far, what we're actually doing is running the Goldie Bench tests for Fable 5 So that'll be interesting to see what comes out in a second. And we'll be updating our leaderboard based on where it gets placed. It'll be interesting to see whether it actually outperforms Fusion, a mixture of agents from Hermes as well.
And also how it performs if you combine it with other models too, if you get even better outputs. So going to be testing out today. The other thing that I've noticed is that we kept getting bugs with Claude today.
It kept saying service was busy every time, even when we tried to use Opus 4.8. So that's something to be aware of too. It's like it might be limited, very limited at first. And also it's only available until July the 7th on the existing subscription. The other final thing that I saw is that people are getting gated. So what happens essentially is they try and use Fable 5 and then it redirects to Opus 4.8 pretty quickly as well. So hopefully that doesn't last too long, but it seems to be a bit of an issue. But if you're wondering what you can build with it, you can see we're building out some games with this and that sort of thing. We'll be testing out more and more. We've got Fable 5 showcase here of some example stuff that we'll build with it. Usually, it looks pretty nice, so we'll see how it performs. It's definitely the most intelligent singular model, but I'm interested to see will it outperform panels of models too. We've got another one here, which is from James Lee. James was asking, how can you link two separate Claude accounts to an Agent Operating System? I've never tested that.
I've never tested that, to be honest, but what you could try is you could run one version of Claude code inside your terminal and then one version inside Claude desktop with separate accounts and try it that way. Or you could have one on a VPS and then one on a separate instance. So two different options. Got a question here from Ilkin. He was saying that basically they want to run OpenClaw and they also want to run Hermes inside the Agent Operating System, but their MacBook is like five years old and not able to handle them locally. So in that situation, if your laptop setup is not good enough, you got a few options. So number one is you could set up a VPS and then link them together. Another option is that you could get a Raspberry Pi. Kevin actually mentioned this too. It's like you got two options there. So VPS or Raspberry Pi, or you could use Tailscale from separate devices as well. For me personally, I run them all locally, but it really depends on your setup. Honestly, what I'd say as well is like these agents are pretty lightweight, so I haven't seen any issues where, you know, you might struggle to set up so far. It's been totally fine. Sam was asking for a Hermes Desktop Masterclass request. So Hermes Desktop is a separate app to the Agent Operating System. We actually have a Masterclass on that already over here. So we have a full video tutorial, step-by-step guide, et cetera, and a 30 day roadmap.
So we've got the full guide already.
But for me personally, I don't use Hermes Desktop that much because I find the Agent Operating System is a lot better. But if you do want a full step-by-step guide, all our prompts, a 30 day plan for implementing it, et cetera, you can get that all here. But I think if I have a choice, you know, if you're watching this, you're like, do I use Hermes or do I use Hermes Desktop or should I use an Agent Operating System? I honestly think the best option here is an Agent Operating System because then you can pull out all the most useful parts. And also you can't really customize the Hermes Desktop in the same way that you would customize your own Agent Operating System. This is pretty wild. So a lot of our members, they use our Agent Operating System, but then they customize it, as I was mentioning before, like exactly how they want it. So what you can see here is this is basically a system, the Robbie setup, where you can basically look through the latest jobs and then there's an approval queue, and this basically does outreach for you. The interesting thing about this is it will analyze, categorize and triage the job opportunities, look through the latest postings, and then forward them into an Agent Operating System dashboard, as you can see right here. Looks really, really cool. It's got like custom message.

6 more minutes of transcript below

Feed this to your agent

Try it now — copy, paste, done:

curl -H "x-api-key: pt_demo" \
  https://spoken.md/transcripts/1000651996090

Works with Claude, ChatGPT, Cursor, and any agent that makes HTTP calls.

From $0.10 per transcript. No subscription. Credits never expire.

Using your own key:

curl -H "x-api-key: YOUR_KEY" \
  https://spoken.md/transcripts/1000775263733