Built by someone who ships
AI systems to production.
Jon Lasley founded Prompt Assay. He has spent 25+ years in enterprise technology, the last several focused on applied AI: production-grade systems built on large language models, multi-agent architectures, and the prompt and Agent Skill design that makes them reliable. He works in prompt engineering, evaluation, and agent design, and stays a hands-on practitioner as the field keeps moving. He built Prompt Assay to give that work a real workbench.
- Experience
- 25+ years in enterprise technology
- Focus
- Applied AI · prompt engineering, Agent Skills, evaluation
- Education
- University of Michigan
- Writes about
- Prompt and context engineering, agents, LLM evaluation
After 25+ years in enterprise technology, I’ve watched a lot of technology waves come and go. AI is different, and the gap between the organizations that figure it out and the ones that don’t is widening fast.
I’ve spent the last several years at the intersection of AI strategy and hands-on implementation: designing production-grade systems built on large language models, building multi-agent architectures, and writing the prompt and Agent Skill systems behind them. I learn this field by shipping in it, not just reading about it.
What I’ve learned doing this work: most organizations don’t have an AI awareness problem, they have an execution gap. The strategy conversations are already happening. The real question is who connects that strategy to something that actually runs.
I’ve been on both sides of that, the person mapping the strategy and the person building the system. That combination is rarer than it should be, and it shapes how I think about tooling. I treat this field as something to keep learning, not something I’ve finished learning.
Building those systems meant writing, testing, and re-testing prompts every day, and the tooling for that work never matched the tooling for the rest of the stack. Prompts lived as raw strings. There was no diff when one changed, no version history, no structured way to tell whether a rewrite actually improved anything.
Prompt Assay is the workbench I wanted for that work: a six-dimension critique on every prompt and Agent Skill, real version history with diff and restore, evaluation suites with LLM-as-a-judge scoring, and a cross-provider Behavioral Eval that runs the same work across Claude, GPT, and Gemini. It is bring-your-own-key at every tier: your keys connect directly to your providers, your inference bill stays with your provider account, and the platform never sits in the request path.
I write about prompt and context engineering, evaluation, and agents on the Prompt Assay blog. The documentation covers the workbench in depth.
Try the workbench.
Free to start. Your keys, your bill, no demo call. The fastest way to understand Prompt Assay is to assay a prompt in it.