Hi, I’m Weller Davis: the writing and project identity of Dave Wheeler, a software engineering leader focused on AI infrastructure, change safety, autonomous systems, and humane technology. I use this site to explore practical engineering, AI-assisted creation, and the systems we need for technology to serve people well.
In my day job, I lead a team building systems for safer, highly automated and increasingly agentic operations at large scale. My independent work through Cephra explores similar themes from the ground up: LLM orchestration, persistent memory, autonomous workflows, MCP tool integration, and AI-assisted engineering.
I’ve written about this work in an essay: The Latency Dividend: Why Agent Safety Belongs Earlier in the Lifecycle.
For professional background, see my LinkedIn profile.
Weller Davis is the name on this byline; the person writing is Dave Wheeler. I write under it so the work can stand on its own rather than become a personal spotlight, not because there’s anything about who I am that I’m keeping from you. My name and day job are right there in the Professional Context section above, and everything below is my own story, just told under the name I chose to write it with.
I started my career at the intersection of people and systems. With formal training in psychology and a deep dive into software engineering, I’ve led teams, shipped products, and built a consulting practice focused on humane, effective technology.
Professionally, I lead the team building safer, agentic operations infrastructure at scale, the subject of my essay The Latency Dividend.
Cephra began as a family assistant: each person gets a private, isolated memory, with an opt-in Circles feature for what gets shared. It already runs my kids’ daily lessons and games on real family data, not a demo.
My latest experiment was Cephra Prime, a local system that consolidates new knowledge into its own weights while it sleeps. Proving it actually learned meant teaching it a fact I invented myself, the Tidepool Protocol, a story I told in The Tidepool Test.
My aim isn’t just to inform: it’s to nudge us toward better ways of working and living together, with tools that are kind to humans.
Cephra is the family assistant behind my household: a chat interface that runs my kids' personalized lessons, general conversation for the rest of us, educational games, scheduled recurring tasks, deep research on demand, and a blog-post generator my wife asked for once and we haven't opened since.
Curious how the content on this site gets made? Mneme is the AI-powered generation engine underneath it: it hosts most of the creators Cephra calls on to build that content, tutorials, blogs, ebooks, code, diagrams, music, and more, all running on consumer hardware.
I've documented that journey while building Mneme in a detailed blog series covering architecture decisions, production challenges, and lessons learned from bringing each creator module to life.
Follow the journey from concept to production across 8+ parts covering ComfyUI migration, music generation with YuE and Stable Audio, code generation, and the architectural patterns that tie it all together.
Start with Part 1: Architecture & Local-FirstWant the technical deep-dive? Download the comprehensive PDF that maps out Mneme's modular architecture, LLM integration patterns, and production deployment strategies.
Download the Field Guide (PDF)The most recent chapter covers the migration from YuE's complex two-stage architecture to Stable Audio Open, a story of engineering trade-offs and knowing when to pivot.
Read Part 5: From YuE to Stable AudioAutonomous AI companies running 24/7, generating daily news, strategic decisions, and investment reports, all powered by the Cephra distributed AI stack on consumer hardware.
Read the deep dive Visit Cephra Daily NewsBy Weller Davis • Apr 01, 2026 • 6 min read
What if I told you there's an AI system running fully autonomous companies, complete with AI CEOs, workers, and daily news...
Read moreClarity over complexity. Fewer moving parts, more understanding.
Practice over perfection. Small experiments beat big promises.
People first. Tech and processes should serve human wellbeing.
Open hands. Share what works so others can build on it.
Occasional notes on psychology, collaboration, and humane tech. No spam. Unsubscribe anytime.
Speaking, collaboration, or media, say hello. I’m happy to connect.
