Writing
Learning out loud
Articles grouped by series, in reading order within each series.
Series
Teaching AI to Read Payer Edits
Payer-specific rules (SNIP levels 6 and 7) are scattered across hundreds of companion guides and bulletins, and most get discovered one rejection at a time. This series follows the pipeline I built to read the guides and turn them into rules a human reviews before any claim is sent.
Teaching AI to Read Payer Edits
Extracting a UnitedHealthcare Smart Edit into structured JSON, and why X12 loop hierarchy is the hard part. Splitting it into layers (and specialized agents)
Payer normalization, pattern recognition, X12 mapping and rule generation as separate problems. Breaking Down the Monolith
One overloaded prompt becomes four sequential steps with independent retries. Is It Really Multi-Agent?
An honest look at the architecture, and why sequential is the right answer for now. The SNIP 6-7 Problem
Why the industry's reactive edit workflow exists, and what a proactive one needs.
Extracting a UnitedHealthcare Smart Edit into structured JSON, and why X12 loop hierarchy is the hard part. Splitting it into layers (and specialized agents)
Payer normalization, pattern recognition, X12 mapping and rule generation as separate problems. Breaking Down the Monolith
One overloaded prompt becomes four sequential steps with independent retries. Is It Really Multi-Agent?
An honest look at the architecture, and why sequential is the right answer for now. The SNIP 6-7 Problem
Why the industry's reactive edit workflow exists, and what a proactive one needs.