Tag
This paper proposes a distribution-aware training approach for modeling next-event predictions in concurrent Go programs, treating scheduler nondeterminism as a signal. Fine-tuning a 7B model on fewer than a thousand traces achieves 36.2% accuracy on production bugs, outperforming Gemini 3.5 Flash zero-shot.
This article argues that using weaker database isolation levels as default is a form of premature optimization, and recommends serializable isolation level unless the DBMS already defaults to it. It cites real examples of concurrency bugs leading to financial losses.