Complex Systems with Patrick McKenzie (patio11) Building software that survives contact with reality, with Will Wilson
55 snips
Sep 4, 2025 Will Wilson, CEO of Antithesis, dives into the future of software testing with a transformative approach focusing on deterministic simulation. He highlights the innovative 'time machines' his team has developed for debugging complex failures in distributed systems. From discussing the limitations of traditional testing methods to illustrating the evolution of testing with insights from classic games like Super Mario, Will emphasizes the need for robust solutions as AI begins to take the reins on coding, shaping the next era of software reliability.
AI Snips
Chapters
Transcript
Episode notes
Distributed Databases Can Hide Complexity
- Distributed databases were once considered impossible to make behave like a single machine, but this view was an overinterpretation of narrow theorems.
- Systems like FoundationDB and Spanner demonstrate you can hide distributed complexity and provide intuitive, consistent semantics.
Physical Reality Scales Fault Exposure
- At large scale the world of atoms matters more: more machines increases the chance of physical failures like cosmic rays or batch defects.
- Correlated physical failures (batch drives, supplier issues) are a central reliability concern for distributed systems.
Complex Software Fails From Unseen Interactions
- Software is an emergently complex system where many failure modes are unforeseeable from design alone.
- Most catastrophic problems arise from unknown interactions and model error rather than single-component mistakes.