Dashboard
Graph

Inspiration

We were inspired by one simple idea: The experience that users have on your site is simultaneously the most important metric and the most difficult thing to measure. With AI as a rapidly developing technology, the user testing process could be automated in a way that enables solo devs/small teams to generate useful data without the previously required resources.

What it does

Friction takes in your website as input (and the codebase if you own it), creates steps for one or more desired user stories, has an agent walk through each of those stories with real time updates, identifies points of “friction” or bugs, and updates the code + creates a PR for human review. User stories can either be generated or defined by the user.

How we built it

We built friction using Creao as the agentic core of the project. Our Creao agent can create steps from user stories, walk through a task using the code/UI, identify bugs and points of “friction”, and make changes to the codebase to fix them. We prototyped Friction’s UI with Paper and followed the Agent Prototyping -> Review -> Iterate -> Translate To Code flow in order to build the frontend in Next.js.

Challenges we ran into

The biggest challenge we ran into was figuring out how to orchestrate our Creao agent in a way that promoted the highest level of AI accuracy while also allowing for our dynamic UI graphs that showcase live progress.

Accomplishments that we're proud of

We are the most proud of our ability to translate a simple idea into a lightweight, fully agentic web app that actually works. We are also very proud of our dynamic, real time UI components that show the agents walking through each user story step-by-step.

What we learned

We learned that, by leveraging agentic AI, user experience data can be generated in a way that closely mimics testing with real users. While Friction is already a working MVP with real results, we also learned that this concept has a ton of room for further development in terms of features and output quality.

What's next for Friction

For Friction to succeed in the future, we will need to test the platform with multiple different models and websites to find gaps and improve AI output accuracy as much as possible. The quality of our findings is and always will be the most important part of our platform.