CI/CD Breaks at AI Speed: Tangle, Graphite Stacks, Pro-Model PR Review — Mikhail Parakhin, Shopify

Shopify CTO Mikhail Parakhin details the company's aggressive AI integration, emphasizing that effective AI engineering requires more than just high token consumption. Success hinges on efficient agentic workflows, such as rigorous critique loops where models verify each other's code to minimize production bugs. The company utilizes internal platforms like Tangle for data processing and Tangent for auto-research, which democratizes machine learning experimentation for non-specialists. Furthermore, SimGym leverages decades of historical sales data to simulate customer behavior, allowing for counterfactual analysis and optimized interventions. These tools collectively enable Shopify to maintain high productivity and quality despite the massive volume of code generated by AI. Parakhin also highlights the shift toward non-transformer architectures like Liquid neural networks for low-latency applications, showcasing a merit-based approach to infrastructure that prioritizes performance and efficiency over standard industry trends.