First month of oneblvd 🚀

Now that June is almost here, roughly a month has passed since the rebrand from BOLT Social to Oneblvd. Tbh, most of the time has been spent laying out the core mission of the platform and what are the immediate/future goals.

The hardest part is honestly staying true to the cause. Every day you hear of a new "AI Startup" raising millions of dollars for x or y. It's enticing for sure. But I do think if everything aligns, oneblvd can be an amazing tool that everyone will use to stay connected, find what they love, and honestly discover the world.

The app also has gone through multiple iterations the last few weeks, from the entire UI design to even specific terms being changed. The final update has just been finalized and it truly does reflect the new vision for the app. Beginning with the dual interface.

The update includes a new text-posting interface, long from captions, the dual interface, and our new and improved on-device model.

In terms of AI progress, the focus currently has been building a sufficient and efficient local model that can get us off the ground. I tend to always lean towards being privacy conscious, so doing all sensitive analysis and computations on-device does help a lot. However, it does have a lot of bottlenecks and consumes far more time than using an existing LLM API. Recently I finished training/fine-tuning the latest version of our model, which significantly improves accuracy compared to the previous editions - mostly in semantic understanding in images. Obviously the performance is not as accurate an OpenAI API or Vertex API would provide, but, with no funding yet, this is the most sustainable and effective way I've thought of.

This month will be focused on rapid app development and the beginning of marketing the app. Hopefully by the end of June we have a larger user-base and can begin working on the true power of oneblvd.

-- Darsh