A developer who openly admits to knowing "nothing about Rust" has successfully created a working PHP engine written entirely in the systems programming language, using AI coding assistants as co-pilots throughout the development process. Ekin Ertaç documented the experiment on their blog, revealing that while they had zero prior experience with Rust, they relied heavily on large language models to bridge the knowledge gap and generate the bulk of the implementation.

The Approach: Letting AI Handle the Learning Curve

Ertaç's strategy was straightforward but ambitious: leverage ChatGPT and Claude to handle both the Rust syntax learning and the PHP engine architecture simultaneously. Rather than spending months becoming proficient in Rust first, they treated the AI models as tutors and code generators, asking questions about idiomatic Rust patterns while pushing forward on core engine components like memory management, string handling, and runtime initialization. The approach mirrors a growing trend among developers who are using LLMs to work outside their primary skill sets, essentially outsourcing the cognitive overhead of learning new languages. The PHP engine needed to handle core language features including variable management, basic arithmetic operations, echo statements, and several built-in functions that WordPress depends on. Ertaç iteratively refined the implementation by running the test suite after each major component landed, allowing AI-generated code to be validated against real PHP-src test cases rather than relying solely on model confidence scores.

Test Results: 17% Passing and Climbing

The current build passes approximately 17% of the official PHP-src test suite—a figure that sounds modest but represents meaningful progress for an experimental project built without manual Rust expertise. More impressively, Ertaç demonstrated that their engine can successfully render WordPress, one of the most widely deployed PHP applications in existence. This practical validation matters more than raw benchmark percentages because it proves the implementation handles real-world code patterns rather than just passing isolated test cases designed for PHP's original C implementation. Ertaç acknowledged significant limitations in the current state: certain string functions remain unimplemented, memory handling differs from PHP's reference implementation, and edge cases around type coercion continue to cause failures. The project exists somewhere between proof-of-concept and production viability—sufficient to spark interest but clearly not ready for critical infrastructure deployments.

Key Takeaways

  • AI coding assistants can enable developers to build complex systems in unfamiliar languages without prior expertise
  • Passing 17% of PHP-src tests while rendering WordPress demonstrates real-world code compatibility, not just synthetic benchmarks
  • The project raises questions about the future role of traditional language learning for working developers

The Bottom Line

This experiment isn't about replacing Rust programmers—it's about demonstrating that AI can compress the skill acquisition timeline dramatically. When a developer with no Rust knowledge can produce a working PHP engine capable of running WordPress in weeks, the economics of software development are shifting in ways we haven't fully processed yet.