What I Built With $0, an AI Agent, and Refusal to Quit

I built something that, on paper, I probably had no business building. I say that plainly, not dramatically. I’m dyslexic. I have ADHD, OCD, and aphantasia. My mind does not move through information in the clean, linear, conventional way that a lot of systems assume it should. I do not come to this from the angle of being a polished engineer or someone with a giant budget or a team behind me. I came to it from obsession, necessity, pattern recognition, and a refusal to let complexity win.

What I wanted to build was simple in concept, even if the execution turned into a maze. I wanted auraboros.ai to do more than just aggregate AI news. I wanted it to metabolize it. Every eight hours, my GitHub cron refreshes the top 10 AI news articles on the site. When those articles change, I wanted the new stories to trigger a complete media pipeline automatically. Summarize the article. Turn the summary into a script. Turn the script into a short-form video. Add text overlays. Generate AI narration. Then distribute those videos automatically across X, LinkedIn, BlueSky, Mastodon, Telegram, TikTok, YouTube Shorts, and Instagram Reels. Not manually. Not occasionally. As a living system.

And I wanted to do all of that for $0.00.

That part matters, because it changes the spirit of the project. This was not a venture-backed experiment. This was not built with a production team, custom software budget, paid voice actors, or premium video infrastructure. This was built using content, an OpenAI Plus account, and an AI automation layer through Hermes Agent. That was the constraint. No extra budget. No glossy safety net. Just whatever intelligence, workarounds, and persistence I could assemble out of the tools I already had access to.

The important thing is that it worked.

Not beautifully. Not at the level I want. But it worked. 

The videos that came out were subpar. They all looked terrible. The pacing was off. The visual quality wasn’t strong enough. The overall feeling was closer to proof-of-concept machinery than polished media. I was genuinely disappointed with the results, because I could already see the gap between what the system did and what it could do if the inputs, rendering quality, narration, and motion language were stronger. But disappointment is not the same thing as failure. In fact, this is exactly where people often misread what they’re looking at.

The success was not that the videos were good. First generation of anything is most likely terrible. The success was that the system completed the loop.

That is the threshold.

The system took refreshed articles, processed them, wrote scripts, generated narration, created videos, and pushed them into a state where automated distribution becomes possible. That is not a theory. That is not a mock-up. That is not “someday.” That is a working chain, even if the output still needs serious refinement. And once a chain works end to end, the question is no longer whether it can be done. The question becomes how to improve it.

That distinction is everything.

Most people wait until something is beautiful before they allow themselves to call it real. I think that’s a mistake. A bad first version that actually runs is far more important than a perfect concept that never leaves your head. A clumsy system that works is more alive than an elegant fantasy. And what I built, even with rough results, is alive.

The deeper reason I did all of this is that I wanted to prove something to myself and to other people at the same time. We are entering a phase where a single individual can orchestrate processes that used to require teams. Research, summarization, scripting, media generation, voiceover, multi-platform distribution. These are no longer isolated skills locked behind departments. They are becoming composable functions inside systems. Once you understand that, your relationship to work starts to change. You stop asking, “How do I do each task manually?” and start asking, “How do I build a loop that can do this repeatedly with me directing it?”

That shift is especially meaningful to me because my life has not been organized around ease. I have had to fight through cognitive friction constantly. Dyslexia changes how I process language. ADHD changes how attention behaves. OCD can turn detail into a vortex. Aphantasia means I do not visualize internally in the way many people assume creativity works. And yet, despite all of that, I have been able to build systems, connect concepts, and push through complexity because I do see patterns. I do see structures. I do see relationships between pieces that other people often miss.

So this is not a pity story. It is the opposite.

It is a story about what becomes possible when you stop worshipping the ideal path and start building from the mind you actually have.

That matters, because a lot of people disqualify themselves too early. They assume they are not technical enough, not organized enough, not educated enough, not focused enough, not resourced enough. They assume that if they do not fit the archetype of the builder, then the frontier is closed to them. I do not believe that. I think the frontier is opening in strange ways, and one of the most powerful things AI can do is create alternate ramps into capability for people who have always been told, directly or indirectly, that they are built wrong for this kind of work.

I am not built wrong for this work. I am built differently for it.

And that difference has become an asset.

Part of why Hermes Agent matters in this story is that it acts like a bridge between intention and execution. It helps absorb complexity that would otherwise splinter the process into too many manual pieces. Instead of having to personally operate every layer with perfect consistency, I can direct the system, adjust the logic, test outputs, and keep iterating. It becomes less about me performing every task and more about me shaping the behavior of an automated process. That is a massive unlock, especially for someone whose cognition doesn’t always map cleanly onto traditional workflows.

The “why” behind all of this is bigger than AI news clips. The deeper motivation is to build an autonomous media organism. Auraboros isn’t just meant to publish articles. It is meant to generate signal, metabolize signal, and distribute signal across the channels where people already live. The top 10 news stories refreshing every eight hours are not just articles on a site. They are triggers. They are fuel. They are input for an increasingly agentic publishing system that can turn information into media without requiring a human to manually push every button.

That’s the real vision.

Not just content creation, but content transmutation.

And the reason I wanted video specifically is simple. Most people are not going to read every article. They are not going to visit a website eight times a day to monitor the latest movement in AI. But they will encounter short-form video on the platforms they already use. If the system can convert the most important AI developments into short, digestible, narrated clips, then auraboros.ai stops being just a destination and becomes a distributed intelligence layer that reaches people wherever they are.

That matters because attention is fragmented. Distribution is everything. And increasingly, the site is not the endpoint. It is the source node.

The frustrating part is that quality still costs. That’s the wall I hit. With a much higher budget, the visual quality could improve dramatically. Better voice synthesis, better motion systems, better templates, better compositing, better scene generation, better timing, better everything. I can already see how incredible this would become with stronger resources. But the fact that I hit a quality ceiling does not erase what was proven. It sharpens the next problem. I am now at the stage of figuring out how to produce these videos at scale, at acceptable quality, for $0.00.

That is a very specific challenge, and it is the kind of challenge I actually like.

Because constraints force design.

If I had unlimited money, the answer would be easy. Pay for better tools. Pay for better rendering. Pay for premium voices and automation infrastructure. But at zero dollars, every choice matters more. Every inefficiency becomes visible. Every weak link gets exposed immediately. That forces a more honest system. You have to ask what can be open source, what can be templated, what can be reused, what can be simplified, what can be done with clever orchestration instead of brute financial force.

And in a strange way, that makes the project more important, not less.

Because if this can be done cheaply, or eventually freely, then it stops being a niche technical trick. It becomes a model other people can use. A student could use it. An artist could use it. A researcher could use it. A small nonprofit could use it. A single person with no team and no funding could use it. That is where this starts to move from being my experiment into being an example of what the general population can actually do now.

That’s the motivational core of it for me. Not “look what I did, please feel sorry for how hard it was.” More like: look at how absurdly open this moment has become. I built an automated AI news video pipeline with with no budget, using an AI agent and stubbornness as the operating system. If I can do this, then the average person is far less trapped than they think they are.

That doesn’t mean it’s easy. It’s not. It’s frustrating, messy, nonlinear, and often demoralizing. I spent real time making ugly outputs. I got disappointed. I hit ceilings. I watched things partially work and knew they weren’t good enough. But none of that changes the central fact that the machine moved. The loop ran. The concept crossed into reality.

And once reality is involved, the whole game changes.

From here, the task is refinement. Better prompts. Better system design. Better ways of generating visuals. Better narration options. Better templating. Better distribution logic. Better use of what is free. Better orchestration between components. The foundation is no longer hypothetical. It exists. Now the work is to make it worthy of the vision.

That’s where I am now.

Not at the beginning. Not at the end. Right in the uncomfortable middle, where something has been proven, but not perfected.

And honestly, that’s one of the most important places to talk from, because too many people only speak after the story is clean. I’d rather speak from inside the mess. From the proof-of-concept stage. From the point where disappointment and possibility are sitting right beside each other.

Because that is where most real things begin.