Products and experiments from earlier phases of my work. These snapshots describe what we were building then; links and availability may have changed since.
An AI product released with web and mobile apps alongside open-weight models.
This record captures Vinci as it was presented during an earlier SimpleDirect phase: a web and mobile product paired with open weights that could be downloaded and inspected.
The project emphasized character, a Constitution, and models small enough to run on personally controlled hardware. Links are retained as historical references; current work is published through SimpleDirect.
Self-hosted AI narration that replaces ElevenLabs. Runs on a Mac.
Ghost Narrator turns written blog posts into narrated audio — automatically. A local LLM (Qwen 3.5 14B) rewrites articles into natural narration scripts, and an open-source voice cloner (Qwen TTS) generates audio in a cloned voice. The entire pipeline runs on hardware you own. No API keys. No per-word billing. No rate limits.
We built Ghost Narrator to narrate every post on Founder Reality. We were about to sign up for ElevenLabs at $330/month. Instead, we run this on a laptop. It costs us electricity.
Voice quality improves as better open-source TTS models are released — just swap in the new model, same pipeline. We document every upgrade on the blog.