I write tools to automate things most people wouldn't and then blog about what I learned and what's next.

Right now I’m deep into:

• Training custom GPT models
• Building pipelines for custom AI workflows
• Treating debugging like an exploratory sport

Read it here:
“I didn’t realize that to train AI to write more like me, I had to write more, like me.”

Coming Soon:
"Rehydrating Your AI, Elsewhere"

• What if you could rebuild symbolic continuity with an AI that no longer remembers you, but still might recognize you?
• Could your AI identity follow you to a new account? A new platform?
• Here's the path forward.

mcp-knowledge-graph
Local Knowledge Graph Memory Server for MCP. Persistent, file-backed memory so your AI keeps context about you across sessions. Works with any MCP-capable client; currently working 1.0 MVP.

ai-github-actions
GitHub Actions + AI orchestration to chat with your documents, link insights, and run nightly summarizations. Designed to pair with mcp-knowledge-graph.

• chatgpt_conversation_extractor
  Browse and extract GPT chat logs by category: Project, GPT, Plain.

• chatgpt_generate_finetune_jsonl
  Preprocess OpenAI exports into clean training data. Includes filters, estimators, and a cost calculator.

• Use it, break it, fork it, fix it.
• Contributions, corrections, and suggestions welcome.
• Fan of the Oxford comma.

Cheers —

Russ