Building with UnaGo: how an AI engineer uses AI to build AI
The day AI quietly moved in
A few years ago, "AI" was something you read about in papers. Today, it's the co-pilot sitting next to me while I write code, the assistant that books my meetings, the engine behind the films I watch on a Sunday, and increasingly, the voice I bounce ideas off before I commit to a decision.
As an AI engineer, I live inside this world professionally, debugging model behaviour, tuning prompts, shipping agents. As a regular user, I lean on it the same way I lean on a good IDE (the software developers use to write and organise code): it's just there, and I'd notice immediately if it wasn't.
The interesting question isn't whether AI is part of my life, it clearly is, but whether I can shape it into something that meaningfully extends what I'm capable of. That's the experiment I've been running with UnaGo, and it's gone better than I expected.
Where UnaGo slots into my workflow
I use UnaGo for a lot of things, but two use cases dominate everything else.
The first is building dashboard apps. Anything I want to see, track, or visualise (internal tools, model performance panels, a quick analytics view for a side project), I knock together in UnaGo rather than reaching for a framework. The iteration loop is short, the data plumbing is already there, and I end up with something I can actually use the same day, not next week.
The second is creative work, and this one matters more to me personally. I've always loved filmmaking. The craft of turning a vague image in my head into something that actually plays on a screen, frame by frame, is the kind of problem I never get tired of. UnaGo let me take that obsession and run with it for video ads and short-form content. The agentic flow for creating videos is straightforward: I describe the scene, the mood, the cuts I want, and the system pulls it together.
Then comes the part I didn't expect to enjoy as much: building video agents that refine those outputs until they match what's in my head. It's one thing to get a video generated; it's another to have an agent that pushes back, iterates, and tightens the result until the imagination actually shows up on screen. Finally, I spin up another agent to measure how the video is performing (watch time, hook rate, the boring but essential stuff), so the creative loop isn't just a feeling, it's data.
It's the closest thing I've found to directing a film where the camera listens.
UnaGo to build a better UnaGo
The other half of my day looks more like classic AI engineering: analysing model behaviour, tuning prompts, reviewing logs, shipping small features. None of that is glamorous, and all of it eats time if you do it by hand. So I wired UnaGo into it.
I now have a dashboard, built in UnaGo, naturally, where I can drop in a behaviour trace, a config diff, or a new feature idea, and use agents to surface what matters. Tweak a parameter here, validate the change there, watch the metric move. It's a feedback loop where the tool I'm improving is the tool I'm using to improve it, which is the kind of recursive elegance I really enjoy. Using UnaGo to make UnaGo better isn't a gimmick; it's the most direct way I know to test whether the platform is actually good.
Creativity has no ceiling, only time
The thing I was most worried about, going in, was the standard worry: does using AI flatten the work? Does it sand off the texture of a human-made video, hand-tuned code, or a design decision I argued with myself about for an hour?
It hasn't. And I think I understand why. UnaGo isn't replacing my taste; it's compressing the time between taste and output. I still make the calls. I still pick the shot, write the line, decide what's good enough. What I don't do any more is spend two days on the mechanical scaffolding that used to sit between idea and result. That scaffolding is gone, and what's left is more of me per hour than I had before.
That's the real pitch, for me. UnaGo doesn't make me redundant. It raises the ceiling on what I can ship in a week. It lets me say yes to more projects, more experiments, more half-mad ideas about a feature or a tool that I would previously have shelved because the cost of producing it was too high.
What UnaGo actually does, and why it works
A few months in, I find myself reaching for UnaGo across more of my week than I expected. Here's what it has quietly become:
Agent orchestration done right. I can chain specialists (research, drafting, review, measurement) without writing glue code every time. The plumbing is already there, and the agents hand context to each other cleanly.
Dashboards and data, in hours, not weeks. Internal tools, analytics views, performance panels, anything visual. The data layer is built in, so I spend my time on the design and the questions, not the wiring.
Creative generation that respects the maker. Video, image, audio, full ad creatives. The control surface is deep enough that my taste stays in the work. I'm not picking from a slot machine; I'm directing.
Real code execution. A Python sandbox, bash, file storage, version history. When an idea needs to become a script, a dataset, or a deployed artifact, it happens inside the same surface I'm already thinking in.
Tool discovery through MCP. UnaGo finds and composes the right tools for a job across integrations (CRM, ads, docs, comms, analytics, the long tail). I describe what I want, and the system picks the path. I don't have to memorise an API catalogue.
Knowledge that compounds. Memory, lessons, conversation recall. The platform remembers what worked, what didn't, and what I prefer. That feedback loop is what turns a tool into a teammate.
Sub-agent delegation. When a task is big enough, I spin up a parallel worker or hand the chat to a specialist. It's the first time "AI assistant" has felt less like a single brain and more like a small team I can actually direct.
Browser automation when I need to verify. Real navigation, real clicks, real screenshots. Useful for QA, research, and the embarrassing number of times I just want a screenshot to prove something actually works.
A workspace I can shape. File tiers (private and shared), tags, search, version history. It's the unglamorous half of a productivity platform, and UnaGo takes it seriously.
What ties all of this together is the simple thing: UnaGo meets me where I am. Whether I'm writing a video brief, debugging a model trace, shipping a dashboard, or thinking out loud about a problem, the surface is the same. I don't context-switch between six tools to do one job. I stay in flow.
Closing thought
I've been on both sides of an AI platform now: the user who just wants it to work, and the engineer who wants to know how it works and how to make it work better. UnaGo is the rare system that respects both. I enjoyed working with it. I enjoyed, perhaps even more, working on it. Contributing to a tool I use every day is a strange privilege, and one that makes the work feel less like maintenance and more like craft.
If you're an engineer or a creator wondering whether to bring an AI platform into your daily loop, my advice is simple: pick a workflow that matters to you, point UnaGo at it, and see whether the tool gets out of your way. For me, it did. Then it started helping me make itself better.
That was the bit I didn't see coming, and it's the reason I'm still here.