What are AI agents and agentic systems, anyway?
A plain-English explanation of AI agents, agentic systems, and what makes them different from chatbots—plus the production lessons people skip.
I write about building and shipping AI applications from three main perspectives, also referred to as Tracks: engineering, business, and product.
Learn how posts are organised — and how to navigate them.
The 3 most important posts to start with, if you haven't already.
Showing 36 posts
Inside PolicyForge's approval workflow: why AI-generated policy recommendations never publish themselves, and how every decision gets recorded.
How PolicyForge generates multiple usable policy drafts instead of one, and keeps your data out of the process.
Inside the design decision that determines which compliance gap PolicyForge shows you first, and why.
Before building PolicyForge, I mapped every failure mode, including GDPR-relevant privacy risks, into a taxonomy that shapes the product.
Why an AI copilot's real safety test isn't preventing every mistake — it's catching and recovering from the ones it can't avoid.
How GovCon Leads Radar turned speed and cost promises into enforceable, measurable guarantees users can hold it to.
How I built GovCon Leads Radar's AI layer to support judgment without ever making the call for the seller.
GovCon Leads Radar's approval flow almost got designed around a feature the product doesn't have. Here's what shipped instead.