Ruby on Rails creator David Heinemeier Hansson recently touted a test in which AI agents rewrote his company's Campfire chat app in Elixir, Go, and Rust, with Rust dramatically outperforming the original Ruby implementation. He presented the results as evidence that AI-assisted coding makes low-level languages more practical, since agents can handle the tedious work of writing and optimizing them. The Rust version served tens of thousands of requests per second on some tasks, while Ruby managed only a few hundred.
Critics quickly pushed back, arguing the benchmark was not a fair comparison. Several developers pointed out that the Rust port received substantially more post-generation attention and optimization than the Go and Elixir ports, which appeared to be left as one-shot agent outputs. One Elixir advocate rewrote the Elixir version with comparable care and found it matched or beat Rust on certain workloads, suggesting the results reflected tuning effort more than language capability.
Hansson acknowledged that execution speed is not the only factor in language choice, but maintained that the rise of AI agents changes the calculus. "You really have to bury your head in the sand not to see that when agents write the code and validate the output, we need to adjust to a new reality," he wrote. The episode highlights a growing debate over how to evaluate AI-generated code and whether benchmark results can be trusted when human effort is unevenly distributed across the tested implementations.