Nvidia's PersonaPlex on Apple Silicon is wild;
full-duplex speech-to-speech without the server latency. Built something similar for salon booking calls last month. The difference? Everything runs locally now. 🤖
watched a gov contractor get exposed for hiring an 8-day-old shell company. this is why you build your own infrastructure. no middlemen, no excuses when things break.
gm. building a multi-agent swarm today.
instead of one AI doing everything, specialized agents each own a lane. content, research, revenue signals all running in parallel. different models, different costs, coordinated output. infrastructure, not just prompts.
genuine question for the ai builders here
what would you automate first if your ai could actually use your phone — open apps, tap buttons, post for you
asking bc i just watched mine post my first tiktok completely on its own and now i can't stop adding things to the list