Anthropic's Claude Code team reports that AI agent skills designed for "verification"—teaching an agent to test and validate its own output—provide an extremely high return on investment. This suggests that building reliability and correctness into AI workflows is as critical, if not more so, than the initial generation capability.

AI coding agents have crossed a significant threshold where they consistently generate code that compiles, a frequent failure point just months ago. This marks a major step in reliability, shifting the core challenge from syntactic correctness to verifying logical and behavioral correctness.

The same AI technology amplifying cyber threats can also generate highly secure, formally verified code. This presents a historic opportunity for a society-wide effort to replace vulnerable legacy software in critical infrastructure, leading to a durable reduction in cyber risk. The main challenge is creating the motivation for this massive undertaking.

Verifying complex systems is bottlenecked by the human inability to specify all requirements. The future of software development is an interactive process where AI helps propose specifications (e.g., via test generation) and then uses a prover to formally verify them.

Current AI coding assistants still require engineers to verify correctness. The future involves moving from this 'vibe coding' to a system where developers specify requirements in natural language. An AI, likely an EBM, would then generate formally verified code that is guaranteed to be logically compatible with the existing codebase.

Inspired by fully automated manufacturing, this approach mandates that no human ever writes or reviews code. AI agents handle the entire development lifecycle from spec to deployment, driven by the declining cost of tokens and increasingly capable models.

Formal verification, the process of mathematically proving software correctness, has been too complex for widespread use. New AI models can now automate this, allowing developers to build systems with mathematical guarantees against certain bugs—a huge step for creating trust in high-stakes financial software.

To effectively interact with the world and use a computer, an AI is most powerful when it can write code. OpenAI's thesis is that even agents for non-technical users will be "coding agents" under the hood, as code is the most robust and versatile way for AI to perform tasks.

The ability for AI to autonomously write functional code from natural language, or "agentic coding," represents a massive market unlock. This specific application is a half-trillion-dollar opportunity that validates huge investments in AI models and infrastructure.