Inspiration
Gimil was inspired by the latest benchmark from mercor the apex-agent benchmark, frontier models fail at tasks because its not that they are intellgent its that they are not complete, its judging a fish on its ability to climb a tree so we took the different approch using a world model to reason through tasks
What it does
Gimil is a forward deployed engineering ai agent, it understands and learns your ai agnet and solutions and your customers workflow and make it so that the intergration is seemless, basically changing an entier workflow into an API
How we built it
we used the foundational model from aule-technologies used daytona sandbox langchain, sentry to build it with daytona computer use it can access tools to do whatever it needs to accomplish a task
Challenges we ran into
the main challenge was intergration of the model to the tools since its a world model we needed a text encoder to help it understand
Accomplishments that we're proud of
we are able to make our agent use the computer to create a custom solution that will be used to actually setup an api from a visual workflow
What we learned
What's next for Gimil
Gimil is going to evolve to be a better faster and more accurate SOTA level engineer.
Built With
- claude
- coderabbit
- daytona
- gemini
- sentry
- zyphar
Log in or sign up for Devpost to join the conversation.