After months of building in private, we're sharing Fantomu publicly for the first time. It's a side project, not a startup. It's not funded. It's just something we've been building because we wanted to see if we could make AI agents that actually work reliably.
Here's the story of how it started, where it is now, and where it might go.
What Is Fantomu?
Fantomu is an AI agent platform. The core idea is simple: agents that work invisibly in the background, handling tasks without you having to think about them.
Think of it like having a skilled assistant who just gets things done. You don't have to micromanage them. You don't have to approve every action. They just work, and you see the results.
That's the goal, anyway. We're not there yet, but we're getting closer.
Why We Started Building
We got tired of managing repetitive tasks manually. We tried existing automation tools, but they all fell short in different ways.
Some were too simple—they could handle basic tasks but broke down when things got complex. Others were too complex—they required a team of engineers to set up and maintain.
We wanted something in between: powerful enough to handle real tasks, but simple enough that we could actually use it without dedicating our lives to it.
So we started building Fantomu. It's been a learning experience, to say the least.
The Learning Curve
We've made a lot of mistakes. We've rebuilt core systems multiple times. We've learned the hard way that building reliable agents is much harder than it looks.
We started by building workflows and calling them agents. They worked, but they weren't actually agentic. We had to throw away months of work and start over.
We tried parsing natural language for decisions. That was a disaster. We learned to use structured outputs instead.
We built systems without proper error handling. They broke constantly. We learned to assume everything will fail and plan accordingly.
Every mistake taught us something. Every failure made the system better. It's been frustrating, but also incredibly educational.
What's Working Well
Some things are working really well:
- Agent execution: The core execution engine is solid. Agents can run tasks reliably.
- Error handling: It's getting better. Agents can retry, adapt, and recover from failures.
- Decision loops: The decision-making is starting to feel intelligent rather than scripted.
- Structured outputs: Using schemas for decisions has made everything more reliable.
These are the foundations, and they're solid. Everything else builds on top of this.
What Still Needs Work
Some things still need significant improvement:
- The UI: It's functional but not polished. It works, but it's not beautiful.
- Documentation: It's sparse. We know how it works, but we haven't written it down well.
- Bugs: There are definitely bugs we haven't found yet. There always are.
- Features: There are a lot of features we want to add, but haven't had time for yet.
But that's okay. It's a side project. We're building it in our spare time, and we're sharing it because maybe someone else will find it useful—or at least interesting.
Why Share It Now?
We're sharing Fantomu now because we think it's reached a point where it might be useful to others. It's not perfect, but it works. It's not polished, but it's functional.
We also want feedback. We want to know what works, what doesn't, and what people actually need. Building in isolation is fine, but building with feedback is better.
And honestly? We're curious to see what happens. Will people find it useful? Will they have ideas we haven't thought of? Will they want to contribute or collaborate?
We don't know. But we're excited to find out.
What's Next?
We'll keep building. We'll keep improving. We'll see where it goes.
If people find it useful, great. We'll keep working on it. If not, that's fine too. We've learned a lot either way, and that's valuable.
We're not trying to build a unicorn startup. We're just trying to build something useful. If it helps a few people automate their workflows, that's a win.
Want to Try It?
If you want to try Fantomu, go ahead. It's available now. If you have feedback, we'd love to hear it. If you find bugs, let us know. If you have ideas, share them.
And if you want to contribute or collaborate, that's always welcome. This is a side project, but it doesn't have to be a solo project.
This is just the beginning. We don't know where it will go, but we're excited to find out.
Let's see what happens.