Multiple commenters question what actual benefits this approach provides over external tool calling, asking for benchmarks, speed comparisons, and concrete use cases beyond elegance.
← Back to Executing programs inside transformers with exponentially faster inference
Commenters remain skeptical about the practical utility of internalizing computation within transformer weights, questioning why this "elegant" approach is superior to leveraging faster, more reliable external tools. Critics highlighted a frustrating lack of benchmarks and released model weights, suggesting that without concrete evidence of speed or training advantages, the project currently feels like a theoretical repackaging of older neurosymbolic ideas. One particularly pointed perspective argued that just as humans outsource complex logic to computers, models should rely on external systems rather than inefficiently simulating internal machines. Ultimately, while the concept holds some curiosity for low-budget experimentation, the consensus is that its real-world value remains largely unproven.
3 comments tagged with this topic