Meta AI with Muse Spark
Meta Superintelligence Lab (MSL) is an org that was newly formed under Alexandr Wang and Nat Friedman. I coincidentally joined Meta around that time and joined the Lab for product development team.
It was a transformative moment where the team departed from traditional software development methodology and jumped into the AI first process where everyone is encouraged to use AI agents to push the code. The whole team gathered in a small conference room in Menlo Park, started to onboard on tooling, and learned how to pull requests.
I was responsible for largely 3 things.
Chain of Thoughts
Muse Spark’s capability was largely unknown, we knew it was a reasoning model, so we needed some representation of reasoning. I built a side panel, and prototyped a lot of different representations that reflect evolving models’ capability. I had to proxy my experience by “it would be like a certain model”.
Local Experience
I co-built the entire front end for local experience largely based on an internal map tool - figuring out how to get the response back from the server, rendering maps, style them, and adding UI. It was a close collaboration with the XR team who supplied map components to different teams. Also very close collaboration internally, since we shared a side panel across the product (shopping, 1P knowledge and more).
Everything was moving very fast, so instead of doing figma mock, I just demoed the experience on my branch. Everything was real except it was just my branch.
Branding
I was part of a small group that explored Meta AI’s new identity. I wasn’t part of the creative group who explored the identity itself, but more on the application side. How it could be stateful, how to be applied to the actual product. I was also responsible for delivering final assets to the web, and various touch points across Meta Apps.
Meta Superintelligence Lab (MSL) is an org that was newly formed under Alexandr Wang and Nat Friedman. I coincidentally joined Meta around that time and joined the Lab for product development team.
It was a transformative moment where the team departed from traditional software development methodology and jumped into the AI first process where everyone is encouraged to use AI agents to push the code. The whole team gathered in a small conference room in Menlo Park, started to onboard on tooling, and learned how to pull requests.
I was responsible for largely 3 things.
Chain of Thoughts
Muse Spark’s capability was largely unknown, we knew it was a reasoning model, so we needed some representation of reasoning. I built a side panel, and prototyped a lot of different representations that reflect evolving models’ capability. I had to proxy my experience by “it would be like a certain model”.
Local Experience
I co-built the entire front end for local experience largely based on an internal map tool - figuring out how to get the response back from the server, rendering maps, style them, and adding UI. It was a close collaboration with the XR team who supplied map components to different teams. Also very close collaboration internally, since we shared a side panel across the product (shopping, 1P knowledge and more).
Everything was moving very fast, so instead of doing figma mock, I just demoed the experience on my branch. Everything was real except it was just my branch.
Branding
I was part of a small group that explored Meta AI’s new identity. I wasn’t part of the creative group who explored the identity itself, but more on the application side. How it could be stateful, how to be applied to the actual product. I was also responsible for delivering final assets to the web, and various touch points across Meta Apps.