Shopify’s AI Phase Transition: 2026 Usage Explosion, Unlimited Opus-4.6 Token Budget, Tangle, Tangent, SimGym — with Mikhail Parakhin, Shopify CTO

In this deep-dive episode of Latent Space: The AI Engineer Podcast, host Sean interviews Mikhail Parakhin, CTO of Shopify, about the company's explosive AI transformation and its strategic shift toward an AI-native engineering culture. Parakhin details how Shopify has achieved near-universal AI tool adoption—approaching 100% of daily active workers—driven by an unlimited Opus-4.6 token budget and a culture that empowers engineers to use any model they prefer. The conversation explores the company’s internal AI infrastructure, including Tangle, a third-generation system for scalable, reproducible ML experimentation with content-based caching and automatic deduplication, and Tangent, an auto-research agent that autonomously optimizes pipelines. Parakhin emphasizes that the real metric isn't token consumption, but the efficiency of high-quality critique loops and the quality of output. He also introduces SimGym, a revolutionary customer simulation platform that uses historical data and multi-model agents to predict conversion outcomes with high accuracy, enabling counterfactual analysis and personalized interventions. The discussion extends to Shopify’s use of non-transformer Liquid AI models for ultra-low-latency search and efficient distillation, highlighting a strategic shift toward hybrid architectures. Parakhin concludes by framing Shopify’s AI stack as a self-improving moat—where more usage leads to better models, creating a powerful network effect that is difficult for competitors to replicate.