>>MeImCo+(OP)
I don't think any of us can know for sure, since the classification and sentiment analysis were both performed by an LLM, and I'm not seeing validation of the results by the author. You can imagine certain security/cryptography-related terms getting coded naively as negative-sentiment by an LLM ("vulnerability", "breach", "flaw").