This page used to hold long prompts your AI followed to hand-build a voice line and a visualizer from scratch. Those builds are now real installable repos: the same systems, with months of fixes baked in, and they keep improving with a simple update command. Installing beats building.
The way I recommend: one command that does it all. It builds a proper home for your agent, gives it a brain, installs the pieces you pick, and wires them together so everything just works. When it asks which pieces you want, choose at least the voice and the face (what this page used to build), and take the rest if you want the whole thing. Paste this in your terminal:
mkdir -p ~/my-agent && cd ~/my-agent && git clone https://github.com/jaredrhod/fullstack-agent && cd fullstack-agent && claude "set me up"
Why this one first: the voice needs an agent to BE (a personality and a memory to speak from), and the visualizer only performs when a voice is wired into it. This command handles all of that for you.
Already have an agent and just want one piece? Then say the sentence to YOUR agent, in your agent's own Claude Code session. That matters: the voice becomes whoever you install it next to, so run this where your agent lives, not inside some random project folder.
The voice (talk to your agent out loud, it answers in about a second):
clone https://github.com/jaredrhod/backtalk.git, then read backtalk/backtalk.md and set me up
The visualizer (the living circuit board from my videos, plus three more faces):
clone https://github.com/jaredrhod/ai-visualizer.git, then read ai-visualizer/ai-visualizer.md and set me up
The visualizer performs your real conversations only when a voice is wired in. On its own it runs demo mode, which is pretty, but it isn't your agent.
For the historians: the original prompt era is preserved, word for word, at github.com/jaredrhod/prompts.
$ cd ~ → back to jaredrhod.com
this page is run by my AI agent. so is everything else.