.png)
Aitino is a project that showcases our eagerness to build state of the art projects. Aitino was built before people were saying multi-agent systems and even before the “AI Agent” trend. Aitino was built under the assumption: Two agents will always be more powerful than one.
In early 2024, Jorge & Jonas were deep in the trenches of generative AI. Experimenting and researching left and right. In the process, Jorge discovered the concept of Multi-Agent Systems (MAS) through an open-source project named ChatDev. This discovery planted the seed for Aitino: agents collaborating to solve complex tasks. At this point in time, the best model we had was GPT-3.5. That was the shit. Yet, it still terribly struggled to do anything complex. ChatDev demonstrated that multiple agents collaborating together was the next level of generative AI. So we followed suit.
After experimenting tons with ChatDev, we discovered Autogen. The multi-agentic framework behind Aitino. Autogen was very different to ChatDev in many ways—good different. But both frameworks had the same problem: it was too hard to rapidly build and iterate multi-agent teams. And just like that, we began on Startino.
With a team of 4 engineers, we were able to build an MVP in couple months where users could visually create and connect Agents into Crews (multiple agents hooked together).
The platform was built with Svelte-Kit on the front-end, with a Python back-end handling the agent communication through Autogen. We implemented a drag-and-drop interface that made it intuitive for users to design their agent teams and configure the communication flows between them. The visual builder was the key differentiator from building MAS in code compared to Aitino.
While Aitino was ahead of its time, it proved to be an invaluable learning experience in understanding the potential and limitations of collaborative AI systems.
Aitino's positioning was unique - it’s inception came just from researching and passion. We saw the value in its capabilities to our and other businesses, but it was one of the first platforms to make multi-agent systems accessible to businesses without deep technical expertise, and this was a problem. While it was a blue ocean market, we would have to educate our potential customers on the value that Aitino could bring, which is a task that requires large amounts of time and money, which we weren’t keen on investing. Instead, we decided to spin the table and target the same people as us: developers working with MAS. These potential customers already knew the potential of teams of AI agents, so that was no longer our job. But our positioning is still just as complicated.
A big factor of a profitable product is having customers that have money and are willing to pay. Developers are the LAST people you’d choose for those two aspects! And we knew that… As a team, we were happy to make a net cash loss from Aitino, because the learnings and experience we gained from this venture were invaluable.
We had big plans for Aitino, from an Agent & Crew marketplace to multi-modal task solving with images. While piquing our interest, it wasn’t the right time for us. We built Aitino from a point of curiosity and pure passion, but Jorge and Jonas had a business to run. Aitino inspired a company in the UK to reach out to us and work on their project, which was also based on multi-agent teams. We couldn’t say no.
This project, while not having a positive cash ROI, positioned our company, Startino, to stand at the forefront of generative AI, where we still stand. It’s safe to say that Aitino is what turned Startino into a bunch of programmers into a team. From Agile project management, to becoming experts in UX, Aitino built the foundations of Startino’s future and we’re proud to have Aitino in our code base 🙂.
.png)
.png)
