Hermes Idea Foundry: Turn 1-Line Ideas into Shipped Apps with AI Agents (Obsidian + Human Approval Gate)
The script demos a new Hermes project manager framework called the Goldie Idea Foundry inside an agent operating system that captures ideas, has agents classify and route them, generates a build plan, and requires only one human approve/reject step before a project manager agent and sub-agents implement and ship real deliverables viewable in a dashboard. Built outputs (e.g., an SEO calculator, a to-do list app, a meditation timer, webpages) can be previewed, rebuilt, and are saved as plain markdown in an Obsidian vault connected to a “memory galaxy” for future context. The presenter also shows a voice-activated “Hermes Jarvis” that can control the computer, discusses using local models or free APIs to manage token costs, contrasts the old idea-to-nothing workflow with the new pipeline, answers a few audience questions, and promotes the AI Profit Boardroom for the full setup, updates, coaching calls, and resources.
00:00 Hermes Idea Pipeline
00:40 Quick Build Demos
01:18 Obsidian Memory Sync
01:57 Human Gate Approval
03:14 Goldie Idea Foundry
04:08 Why Ideas Die
04:58 Live Q&A Updates
05:45 Hermes Jarvis Voice
06:22 Five Step Workflow
07:54 Old Way vs New
09:31 Proof It Ships
09:49 Board and Ownership
10:41 Costs and Local Models
10:56 Offer and Beliefs
11:49 Member Results
12:22 Models and Recap
14:27 Full System Tour
15:05 Community and Closing
Full transcript
This new Hermes Project Manager is absolutely insane. So what we've got here is a system basically where we can capture our ideas from our AI agents, then we get a human approval, from there we can implement stuff and then we can actually see what we've built directly. So you can see an example of what we've built and the notes of what was built in the meantime as well. And so the great thing about this is we can basically take an idea, capture it, implement it and then get started from there.
And you can see all this cool stuff that we've built out directly using these AI agents. Again like just loads of cool stuff here. We can actually click that and preview it as well in terms of what we've built and then we can see the notes previously as well. Here's a number of examples.
This was a to-do list app. If you have any ideas you can just plug it into the system and then the AI agent will plan it for you, start building it out. So let me show you an example. So if we drop an idea here and we're like okay I need an SEO calculator, something like that, just totally random.
So we plug that in, we capture the idea. From there we're going to shape it. So we click on shape it and then it'll actually ask us to approve it. Once we've done that we can see what was built, what was created, that AI agents just go off and build it.
So there's only really two things here. We give it an idea, we approve the plan and then it goes off and implements everything for us, which is pretty crazy when you think about it. And so you can capture an idea, agents classify and route it, plan it, you approve it once and a project manager and sub-agents actually build it and then it lives in your obsidian vault. And this is the big thing as well.
So everything that we do here is connected to our memory galaxy with obsidian. You can see it right here. And then anything that we do, here's an example, anything that we do like this is plugged into our system. So we can see what was created, what was built, as you can see right here.
And so we have the notes and then our AI agents can use that as context as well in the future. So if we go back to the pipeline section here, we can start creating stuff. So this is a meditation timer app. As you can see that we've just plugged in as a project and now it's beginning to shape and plan it.
From there, inside the human gate, we can click on this. This is a plan that the agent has just created in real time and we've got the idea, the classification, the proposed plan in terms of building it, the approach in terms of how you actually do it, the first milestones and who builds it. So you can see who's going to build it right here and who's going to be assigned to it. So if we click approve, then it's going to go off and do it.
If we click reject, then it's not going to do it. So if we click approve here, that will now start working on this project in the meantime and just start running the background. And so you can have your AI agents working 24-7 to build new cool stuff out and just constantly create new useful things that would actually help you. So you can see here it's now in the building stage, right.
So it's just moved from the human approval section to the implement section and then we can see that we've got the whole setup here in terms of the notes, the plan, the approach, whatever. Now once that's actually completed, you'll be able to preview it like you can see right here. You'll be able to see what was built, how it was created, etc. It's just running in the background for us.
So it's a team of AI agents just constantly creating and building together, which is pretty amazing when you think of it. And so this is a framework I call the Goldie Idea Foundry and it's a new framework that I've just built into our agent operating system. So you can drop an idea and you can see another example of what was built right here. So we can see the example what was created and built, etc.
But you drop in an idea, any idea, it could be half-formed, it could be one line, it could be a voice note, whatever. The agents take over, they figure out what it is, they route it, they write a plan and there's just one place where you step in, which is where you approve or reject it inside the human section. So again, let's try another one. This is the to-do list app that's being planned out now and so they're routing it, they're writing a plan, there's one place where you step in and then you've got a project manager agent and sub-agents building this stuff out for us, which is crazy stuff.
So you can see this is now ready to build and ready to go and same with this calculator idea so we can shape that out and start building out and the thing and the reason that I did this is because in the old day and I'll just approve this calculator example so if we click on approve there we'll start building it. In the old day I had like loads of ideas and every idea hit the same wall, which is I'd write it down and then nothing would happen and there was no time to research it, no time to plan it, no time to build it and so the idea just fell into the background and never got implemented. So by the time I came back I'd forgotten why it mattered and so this is why I actually created the idea foundry into my agent operating systems. Now I can drop in an idea, walk away, my agents classify it, plan it and hand me one decision, I tap approve and then by the time I'm back the thing is built and running right.
So the ideas they don't just disappear, they actually get created in the same day as you can see right here and so I want to talk you through the whole process. Let's see what questions we've got here. Scott says you are AI right? No I'm a human and then here's another one.
What do you think of the new skill github repo that will help agents search and install the skill needed? It could be useful but you've got to be careful what you install if you're creating external skills. I don't really install that many external skills just to play it safe because that was where all the problems came with the setup and then Saber says is Jarvis already had added to the agent operating system? It is, we actually created a new update this morning with Hermes Jarvis and the idea foundry and everything else.
You can get it inside the agent operating system here so you can see the last update date inside the airport for boarding and then if you scroll down you can see in the resources section here. So this is a brand new update with the Hermes section and the Hermes Jarvis section. If you're watching this you're like what is Hermes Jarvis? That sounds cool.
So basically what we've done is we've created a voice activated version of Hermes that can actually control my computer. Now it's playing sound right here. I'm not going to play it out loud because I don't think it's going to sound that good on this video but that'll basically run in the background operating and then we can actually plug in a wake word so that we can get it working and you can see it's actually replying and then it's taking what I say right now and then using that right. So it works with everything else inside the system and it's a super powerful way to just control your AI agents with Hermes Jarvis.
So back to the pipeline. The way this works is there's five simple steps. So step number one is the capture. So you drop the idea and forget it right.
No structure, no tags, no folder, it could be a voice note, a link, it could be one messy line anywhere. It lands in the vault and the agents take it from there. So the friction that used to stop you is gone and we can just plug that in. Then we have the classifying root section.
So an agent reads the idea, decides what actually is a project to build, a quick task for example, an idea to park, a reference to keep or something to flag for you and the right ideas get pushed forward, the noise gets filed and you sort nothing by hand. So you can see for example here this one it didn't build because it's like that's not that useful but then for example for everything else that this SEO calculator tool actually created it and we can preview it we can actually open it up in a new tab as well and then we can comment on it we can click rebuild as well so if we want it to go again and have another shot of that we can get it to do it there as well from there for anything worth building an agent drafts a real plan what it is the approach the milestones in which sub-agents will build it so you get a finished proposal to react to not a blank page to fill and so as an example here you can see the stuff that is created and planned out there then we've got the one gate so this is the only place you step in basically inside this section you just approve or reject the ideas that's all you do and then the forge section is like on approval a project manager spins up his sub-agents and builds the actual deliverable which could be like a working landing page an app a game it ships it you watch it appear running inside your dashboard and so if you look at this the old way you'd be writing the idea in a notes app you mean to research it this weekend open up a blank doc stare at it and then close it because you don't have time to focus and then you forget why the idea mattered you get a freelancer to try and do something they take two weeks most ideas never get built and the result is just tons of ideas you don't actually use or implement the new way is that you drop an idea into the idea foundry you got one line no structure agents classify and route it for you an agent drafts a full plan you approve it once from anywhere sub-agents build the real deliverable it gets shipped running in your dashboard and the result is ideas actually become shipped things so as an example of this if we look at the calculator tool we can see that's built great that was and also we can come back to this so we go back to let's say we refresh a page or we go back to goals and then we go back to pipeline here we can see this stuff we've built it's all saved in one place which is great for example if you were using imagine you were using hermes inside the terminal or even inside the desktop app i think it'd be pretty difficult to do this but if you use it inside the terminal like this right how easy how difficult is it to see everything you've created and get ideas done very difficult whereas for example if you do it in this way you can see everything in one place and you can organize it it's way better saber says can you add office visualization when agents do work in an agent group chat do you know what i'm gonna write those things down that would be love that idea thank you so much i'll definitely work on those channel says i'm coding a hermes agent right now nice that's awesome yeah we're just making it more and more powerful every single day and so that was the old way new ways you drop an idea agents just go off and do it that's it now somebody you want to say this sounds like a gimmick is actually building real things so you can see they're actually creating real stuff right and literally just one idea so that's basically how it's done you see another example of a web page we created over here if you want to preview it in full screen we just click on that button and then we can open up full screen and see what we've created which is pretty awesome and so you got one board left to right from inbox to shipped so this is the inbox where you capture stuff and then it actually gets created over here which is pretty amazing so everything just sits on one board flowing left you can watch your ideas move through it you drop a line into the capture bar at the top it appears in capture you tap shape it in about 15 seconds the agents classify it and then it slides into the human game you read the plan approve it and then also the cool thing is everything gets saved into your obsidian vault there's a plain markdown so your whole idea pipeline is yours and you can search for it it's readable you can come back to it later etc here are some examples of what we've built right so you can see some examples of what we've created what we've built with it etc pretty cool and yeah and we can come back to it anytime as well it's pretty amazing and so here's an example of something else that we built with this ai agent and this can just run 24 7 without you just building out stuff which is pretty cool as well now if people say this is going to use a lot of tokens blah blah blah but if you set it up with a local model if you set up with free api you don't need to worry about that and then you've got everything in one place so you've got by the way this is all part of the agent operating system the idea foundry the full operating system i built that connects claude open core hermes into one dashboard with the full zip file the obsidian memory coaching course where we can set up a 30-day roadmap as well to implement it you can get that all inside the ai profitable room link in the comments description or go back to profitable.com now three beliefs here holding you back if you've watched this far and still haven't committed to building this there's probably three beliefs in the way so let me handle each one number one is some people say my ideas need me to build it right i ai can't really do it but you just saw five real builds from five sentences so the agents do the building you just do the approving other people say i'll set up a system like this one out of time but you set up once and then it gives you time back forever so every idea ships without you doing the work other people say i'll wait i'll wait until this is fully ready and ready to go but the people building with ai agents right now are going to be way ahead when everything gets better so every idea you create now is going to make a huge difference and if if you're wondering okay what are other people's experience with the agent operating system you can see nick is crushing it ethan he said i'm 61 years old i'm not a coder but i have the agent operating system set up and i have eight agents and voice and obsidian installed and he said he's so proud of himself and the others who are actually doing this stuff if nick can do it if ethan can do it if he's inspiring people can do it you can do it too right these are just normal people they're not developers or coders they're just people getting awesome results because they actually implement this stuff so these are all air profitable members by the way let's have a look what questions we got here are all your agents running for free or with the alpha api or do you have local models we have local models running sometimes for me personally i like to mix it up right so i'll test local model you'll see this on my tutorials like i'll test local models i'll test new models test free apis and just show you what works and what works right now but this stuff always changes like even day today for example if you look at seven days ago nemetron 3 ultra was it free on hermes no is it free on hermes now yes this stuff is changing all the time and that's why we change our models all the time and see so basically to do this all you do is you open the pipeline tab drop an idea shape it read the plan and approve it watch it build and then you can view what was built later as well as you can see here and also if you look at the background this is connected to our obsidian memory galaxy which is basically your mind visualized into a beautiful system as you can see and so just to recap here in terms of what you have learned you've learned how to stop losing ideas so you can just drop it once and the agents build it for you you stopped having to sort and plan everything because the agents classify and root everything for you you stopped planning from scratch because you just react to a finished plan not a blank page you also make one call so you can just approve or reject this stuff and then it gets built you stopped building by hand because sub-agents build the real deliverable and actually creates it and this can run on local models we ran this on local models which is pretty cool as well and then you've just got one human checkpoint and everything else is your ai agents so if you wanted to build something like that out you could build it for free but if you want to get my setup it's the full agent operating system with the vault the local models etc you can get that inside the airport for boarding you know you skip all the figuring out and you can just grab the full system from me the foundry the agents the memory all wired together every new piece i build as well makes the whole thing more powerful and we improve every single day so you get the agent operating system zip file every prompt the obsidian memory setup so your agents know everything about you your four coaching calls a week where we can set up together a 30-day roadmap daily tutorials a prompt library and a member map and all of my new setups for this stuff so the agent operating system as you can see is just it's being built into something that's absolutely mind-blowing we've got hermes jarvis in there we have all of our sessions say we have goal mode workspace mcps we've got mission control here and we've done that for all of our agents look at claude right if you go to claude here you can see that we've got this awesome system set up where we can have teams of agents running as well which is pretty insane when you think about a video agent we have an seo agent we have the ideas foundry and we have the memory galaxy as well if you like teams of agents you probably like the hermes paperclip setup that we've got over here as well so feel free to get it link in the comments description or go to the aiprofitboarding.com this is an awesome community you can ask questions get help and support and you'll see loads of people setting up their agent operating system and getting awesome results inside the calendar you can jump on weekly coaching calls inside the classroom you can get access to all my new trainings we have a full beginner to expert course here if you're a complete beginner and if you want to get new daily updates and advanced stuff and new stuff you can get that inside this section we have the agent operating system with a video tutorial you can see when it was last updated so we update this daily and you have the zip file for implementing this all as well so feel free to get that you also get four weekly coaching calls and a map where you can connect with people in your local city we're using ai agents just like you so feel free to get that link in the comments description or go to the aiprofitboarding.com thanks for watching
More episodes