REPL-driven development is often treated as a local expression-evaluation mechanism. But when production incidents span complex business flows, we. abandon our flow state for slow, diagnostics. What if your architecture could automatically materialize individual production incident profiles directly into your local REPL?

This talk explores an advanced debugging workflow that scales conversational programming across multi-service business states. We explore the architecture of an automated pipeline that queries a live, failing entity's history, attributes, and generates a local scratch namespace file populated with real parameters, commands, and events. Using state-flow monadic testing and local variable-capture macros, developers can interactively step through historical transitions and hot-patch code. Finally, we show how this single-entity precision unlocks a massive future: using AI assistants to safely execute and scale these REPL flows across dozens of edge cases simultaneously.

Juliana Prieto

Juliana Prieto is a software engineer at Nu, working with Clojure on backend financial systems. She enjoys learning deeply, teaching others, and communicating technical ideas clearly. While she has a strong interest in frontend, product, and user experience, her current work is focused on backend systems, debugging workflows, and real-world business flows. She is especially curious about software architecture, feedback loops, and the way engineering practices shape how teams understand, change, and trust complex systems.