Why I Started Building RuleKiln
RuleKiln started when a stronger model generated better prompts from examples than manual prompt tuning. This post covers that origin and the first distillation loop behind the project.
Read more →Blog
Insights on AI consulting, process automation, and agentic workflows
How I Turned an Old Mining Rig into a Local LLM Server with llama.cpp, Vulkan, and Three RX 5700 XTs
I bought an older mining rig with three RX 5700 XTs and low RAM, upgraded it, and turned it into a queued multi-GPU local inference box using llama.cpp, Vulkan, Docker Compose, and HAProxy.
Read more →I Rebuilt Product Pages for AEO as a Solo Founder — Canonical URLs, Category Hubs, and Durability Proof
I spent the past week tightening canonical URLs, adding crawlable category hubs, and making durability proof a first-class part of every product page on OnlyBuyForLife. Here's what I changed, what...
Read more →From Reddit Threads to Product Listings: OnlyBuyForLife's Ingestion Stack
We built a product discovery engine that pulls Reddit community wisdom through an LLM, enriches it with live eBay listings, and serves it to shoppers. Here is what broke along the way—and how...
Read more →Data Wrangling in an Ever-Changing SaaS System
How moving from on-prem SQL to SaaS APIs broke our data pipelines, and how we used Python and Pydantic to build a resilient, version-tolerant ingestion system.
Read more →