AI Language Goose Claimed to Beat Rust, but Its Benchmarks Were Rigged
The coding community loves drama, especially when an experimental language claims to destroy established titans like Rust and C++ while leaning on automated generation and unorthodox memory tricks.
The hype started when the experimental language Goose surfaced on GitHub, touting benchmarks where it drastically outperformed Rust in execution speed and slashed memory usage. Skeptics immediately tore into the code to investigate whether compilers had truly been rendered obsolete overnight.
A deep dive into the repository revealed that the comparison relied on a heavily skewed contest. The benchmark pitted an optimized Goose program against an intentionally clumsy Rust implementation that performed dynamic heap allocations on every single string, naturally tanking performance.
When independent developers rewrote the Rust benchmark to store fixed strings inline without heap calls, the massive speed gap instantly vanished into a dead tie of roughly 46 milliseconds. After the critique gained traction online, the repository owner quietly deleted the disputed fixed-mode benchmark file altogether.
Yet underneath the questionable marketing lies a genuinely radical architectural experiment. Goose completely eliminates traditional garbage collectors and manual heap allocation in favor of compile-time data stacks, where memory is reserved upfront and reclaimed like a stack as scopes exit.
By packing variable-length nested data sequentially without heap pointers, Goose serialized complex order objects in just 29 bytes with zero allocations, compared to over 150 bytes and multiple heap hits in typical safe systems code.
Raw compute parity remains out of reach because Goose still lacks multithreaded loop splitting and vectorized SIMD execution across multiple hardware cores.
Novel runtime models often deliver stunning single-threaded memory numbers right up until real-world software demands multi-core scaling, concurrency, and recursive data flow.
Source: Goose
Comments
This is where the magic happens: AI reads your discussion and rewrites the article based on the most interesting comments. Each strong comment adds points to the meter below. Once the meter is full, the article updates live — no page reload needed.