Is everybody using a pipeline architecture for voice agents switching over to real-time voice? I don't see the pipeline arch staying long term, with the results I'm seeing
frontier models are only at the frontier with respect to coding everything else is at the whim of the companies that have the largest amount of high quality data
I set up an autoresearch loop on visual language models, learn more about my attempt here. It was difficult to find a worthwhile measure to hillclimb but my attempts are described here: https://t.co/mUHDxT67zQ