Jan 2, 2019, 12:35 AM ET

a16z Podcast | It's Complicated

Back-end systems will remain complicated while front-end interfaces stay deceptively simple, creating a persistent design tension that evolves into "systems that nobody understands" as AI, deep learning, and probabilistic reasoning replace deterministic, procedural logic.
Future systems, including self-driving cars and mobile apps, will function as self-learning, combinatorial entities where individual components fail continuously, requiring engineers to adopt a "biological mindset" focused on subsystem robustness rather than grand, failure-free equations.
The industry will shift toward managing complexity through group vetting mechanisms like Change Review Boards and iterative testing tools such as Netflix's Chaos Monkey, acknowledging that no single individual, including CTOs or architects, can grasp the full scope of modern infrastructure.
Users and engineers will face a growing gap between mental models and actual system operations, leading to inevitable bugs and unexpected behaviors that necessitate a cultural shift toward humility, acceptance of occasional failures, and a focus on constant iterative tinkering rather than full comprehensibility.
Design and verification processes must fundamentally transform to handle probabilistic systems where decisions are pattern-based, and inventors will increasingly rely on accumulated group expertise rather than individual genius to navigate non-linear, large-scale architectures.