Independent AI Product Builder & Operator
About George O'Nair
Field Lab focuses on one Builder task: assessing a local repository and packaging an authorized workflow as an Agent Skill. Other standalone tools remain in the shop for codebase audits, spreadsheets, small-business operations, and presentation work.
The Field Lab
The internet is full of AI demos that look miraculous for 30 seconds, then collapse the moment they encounter real client constraints, messy tables, or human deadlines.
I run this field lab to build, test, and operate practical AI systems. I test coding agents against messy codebases, build automated layout gates, and run workflows until they break. When an experiment yields an artifact reliable enough to stake client reputations on, it gets cleaned up and packaged here.
Five Operating Principles
01 / BUILDER-FIRSTReal tools, real code, real files. No buzzwords, no thought leadership without proof of work. Every observation comes from sitting in the terminal building things.
02 / SKEPTICAL BY DEFAULTNot an AI cheerleader. I test new models on real tasks and publish where they fail. If a workflow requires four retries to produce one usable slide, it is not production-ready.
03 / SURVIVES REALITYA slide deck or code tool must work under pressure. That means strict layout gates, system-safe typography, and zero dependency on a proprietary cloud viewer that crashes during a presentation.
04 / HONEST CONSTRAINTSEvery artifact clearly states who it is for and who it is NOT for. If a tool requires running commands in a terminal, we say so upfront rather than pretending anyone can use it in 2 clicks.
05 / ZERO HOSTAGE INFRASTRUCTUREThe listed products use one-time plans, not monthly subscriptions. Checkout and account-based access run through Whop; each product license defines how its files or scripts may be used.
Where to Follow the Work
Work in progress is posted on these channels: