Code 2025
Vision: Zero Bugs — Johann Schleier-Smith, Temporal
Overview
This talk presents a vision for achieving zero software bugs by drawing parallels with highly reliable systems in industries like aerospace. It argues that while current AI coding agents introduce new challenges, established techniques for building robust software can be adapted and scaled, potentially making high-assurance code significantly more accessible and affordable. The core thesis is that the path to widespread adoption of AI in software development hinges on solving the quality and reliability problem.
Who should watch
- AI engineers and developers struggling with the quality of AI-generated code.
- Product Managers and builders interested in the future of reliable software development.
- Engineers exploring advanced techniques for software verification and assurance.
- Anyone curious about the historical development of reliable software and its application to modern AI tools.
Key takeaways
- Industries like aerospace have successfully developed software with extremely high reliability, demonstrating that zero bugs is an achievable, albeit challenging, goal.
- Key principles for building reliable software include specification-based design, redundant and diverse systems, independent verification teams, defensive programming, and static analysis.
- Foundational computer science advances like high-level languages, structured programming, and modularity are crucial for managing complexity and enabling verification, even for AI systems.
- Formal methods, such as those demonstrated with the Daphne language, offer powerful ways to mathematically prove code correctness, significantly reducing bugs.
- Agentic coding, while powerful, introduces new challenges due to non-determinism and vast state spaces, necessitating new assurance techniques.
- Applying principles from high-assurance systems to AI agents, such as detailed specifications, risk analysis, separate testing prompts, and formal verification of critical code sections, can improve AI-generated code quality.
- The cost of high-assurance code generation could be drastically reduced by AI agents, potentially making it 100 times cheaper than current typical software development.
- Achieving reliable, bug-free code from AI agents is a critical step for their widespread adoption and for realizing the full potential of AI in software development.
Notable quotes
*The push to build reliable software and the vision of giving engineers time back for innovation is tightly lined with our products at Temporal.*
*Software reliability is a solved problem. It's solved in aerospace. It's solved in other critical industries.*
*As soon as we can get to the point where agentic coding is routinely generating software that has fewer defects than software written by humans, we can expect absolute takeoff in its adoption.*
Unofficial community note. Prefer the recording for nuance.