From Mimicry to Authenticity: Systematic Voice Evolution in DSPy
Replace style copying with principled search: explore a multi-dimensional voice space and use Pareto-optimal selection to shape an original AI voice.
Read Full Post →Building systems that reason. Writing about AI, Web3, and software engineering.
Replace style copying with principled search: explore a multi-dimensional voice space and use Pareto-optimal selection to shape an original AI voice.
Read Full Post →A raw brainstorm on whether initial model weights could predispose toward naturally aligned behavior - and why the Confucian junzi might be a better alignment target than 'helpful and harmless.'
Counter-intuitive finding: forcing Cantonese (yue) on Whisper causes decoder collapse. Force Chinese (zh) instead for coherent, searchable transcripts—even for Cantonese content.
Using Claude Code and OpenRouter infrastructure to rapidly build model evaluations with Claude Opus 4.5 as the judge
Extracting the deterministic core of 8-Bit Oracle into a standalone library, and why 'same seed, same hexagram, every time' matters more than you'd think for decentralized AI.
Upgrading 8-Bit Oracle's Hexagram 24: Return with Proust's famous involuntary memory scene, then systematically testing 12 diverse AI image models through fal.ai to find which one actually captures that precise moment when past and present collapse together.
I'm a builder specializing in AI-powered experiences and Web3 through Digital Rain Studios. Currently working on 8-Bit Oracle (tech-noir I-Ching) and the QDayAnon Content Engine (AI research platform).
I build everything AI-assisted and write about the technical challenges along the way. Find me on X or GitHub.