I tried using code-supernova-1-million as a cheap QA that would increase test coverage and fix problems in my TypeScript project along the way.
As a result, I had 207 tests, of which 187 were passed. I switched to gpt-5-high (I also wrote “gpt-5, I’m switching to you”) and asked to double-check changed tests, as well as finish the work; at the end of the prompt, I summarized the context via /summarize.
After completing his work, I asked prompt:
Can the quality of the tests written before switching to you be qualified as good?
Just answer the question.
Full answer of gpt-5-high
No.
