Thoughtworks Technology Podcast Context engineering: Tackling legacy systems with generative AI
42 snips
Aug 21, 2025 Birgitta Böckeler, a distinguished engineer and generative AI expert, and Chandirasekar Thiagarajan, a seasoned tech principal with retail experience, dive into legacy modernization. They discuss how generative AI can reverse engineer aging systems without full access to code, revolutionizing the process. The conversation includes practical applications of tools like Playwright for user journey analysis and data-driven insights to reconstruct application logic. They also highlight challenges of decompiling legacy binaries and the need for careful risk management when using AI.
AI Snips
Chapters
Transcript
Episode notes
Two-Step AI-Driven Modernization
- Generative AI works best as a stepwise assistant: first describe existing behavior, then use that description for forward engineering.
- Detailed textual descriptions enable AI to generate better replacement designs than direct code translation.
Automate UI Discovery With Playwright
- Use automated UI exploration (Playwright MCP) to capture DOM and user journeys for specifications.
- Feed those captures to AI to generate detailed functional descriptions and screenshots automatically.
Iteratively Build A Master Spec
- Iteratively enrich specifications by running different journeys and merging outputs.
- Ask AI to merge and reconcile documents to create a master specification quickly.