Jev in production › Code and security review
Jev scores a fact sheet for every on-chain transaction to flag hacks; vendor benchmark claims 56 of 60 known hacks caught (company).
Jevscan scans every transaction in a block for possible smart-contract exploits. It reconstructs what happened, derives the economic facts, and asks Jev, a language-model classifier, whether those facts describe an attack.
In our benchmarking, Jevscan caught 56 of 60 known hacks (93.3%). Across 36,357 transactions, it raised 21 potential false positives (0.058%).
Scanning a block took an average of 2.7 seconds with supporting data, such as contract history and historical prices, pre-cached. The block itself, its receipts and call traces were fetched during processing. See the benchmark results and methodology.
For the project's own README, linking back here: