Almantas Karpavičius explores how the shift toward agentic development has flipped the bottleneck: teams now spend more time reviewing AI-generated code than they ever spent writing it manually. When an LLM becomes the primary author of a codebase, he questions whether traditional clean code principles accelerate the agent or simply increase token overhead.
Almantas presents a comparative series of agentic workflows used to build Arcanum Audio - a multi-track audio mixer mobile app. Attendees can expect to see multiple prototypes implemented using different workflows.
In this session, attendees will:

Quantify the reliability gap by comparing no-test, TDD, and BDD workflows using objective metrics such as build failures, features completed in one attempt, bugs, manual interventions, tokens used, context window size-to-quality ratio, and static analysis.

Master agentic economics by learning how to manage token costs on a one-man-army budget while maintaining high-quality output.

Audit the agentic stack by exploring how the choice of tools and models determines outcomes.

Whether attendees are testers wondering if BDD is finally cool again, developers concerned about their craft, or leaders looking to scale output without increasing technical debt, they will see a series of live demos, concrete performance metrics, and the “bruised” results of building real software.