Is Prompt Engineering Dead? How Auto-Optimization is Changing the Game

Explore the evolution of prompt engineering in this 30-minute conference talk from EuroPython 2025 that examines whether manual prompt tuning is becoming obsolete in the age of automatic optimization. Discover how the challenges of time-consuming, model-specific prompt engineering are being addressed through data-driven approaches that can automatically learn optimal prompts. Learn about various prompt optimization techniques, starting with simpler methods like bootstrapped few-shot that automatically generates and selects optimal demonstrations for LLM chains, then progressing to more sophisticated approaches such as MIPRO and TextGrad that directly optimize instructions. Gain practical insights into implementing these techniques using popular frameworks like DSPy and AdalFlow, complete with hands-on demonstrations. Understand the trade-offs between manual and automatic approaches in terms of costs, complexity, and performance to make informed decisions about your prompt engineering strategy. This presentation is designed for professionals working with large language models and assumes some familiarity with the domain, offering key insights into the challenges of manual prompt engineering and the potential solutions provided by automatic optimization techniques.