When Difficult Tasks Are No Longer Difficult

Using LLMs, formal soundness proofs for a new type system in the Move language were completed in just 4 weeks. This is a groundbreaking case of drastically shortening the mechanization of metatheory, which previously accounted for 80-90% of the total effort in Programming Language (PL) research.

This shift is fundamentally changing the publication culture in PL research. In the past, massive human effort (large-scale implementation, benchmarking, manual proofs) was essential to gain recognition for research value, but now skilled researchers can produce high-quality papers within a month using LLMs.

Indeed, changes are already visible, with the number of POPL paper submissions nearly doubling from around 350 to 600. This is not merely an increase in AI-generated content, but the result of automating tedious processes like proofs and implementations, thereby increasing research speed by 10x.

Going forward, PL researchers must set higher standards. Beyond simply defining computational models, they should focus on more ambitious challenges that were previously unimaginable, leveraging modern tools like LLMs.