Auto-graded unit tests are commonplace in programming education, yet their use as pre-learning diagnostics remains underexplored. This study evaluates a short, optional, pre-course Python diagnostic built around unit tests and immediate pass/fail signals. Although the surface task (imple-menting add(a: int, b: str) -> str) appears simple, the suite demands robust handling of messy numeric inputs (locale formats, fractions, scientific nota-tion, bases, complex numbers, currency symbols, verbal numerals). The in-vestigation addressed three questions: (i) how instant pass/fail affects mo-tivation, confidence, and self-esteem; (ii) how students judge usefulness, fairness, and transparency; and (iii) which design changes they prefer.

Engagement logs from students (N=18) and post-course interviews (N=15) were analysed. One third completed the full suite, while many non-completers still reached advanced cases, indicating exploration beyond su-perficial tinkering. Interviews showed that the test felt authentic and moti-vating when it resembled real development work; however, repeated opaque failures reduced confidence. Participants reported that usefulness and fair-ness would be enhanced by an immediate linkage of tests outcomes to con-crete study steps and by the provision of a minimal, autonomy-supportive hint after repeated failure. Changes in intrinsic motivation followed a con-sistent pattern: curiosity at first ‘fail’, a short confidence boost at first ‘pass’, and a dip during repeated failure that could be recovered with a brief informational cue.

Two complementary typologies are proposed to interpret these responses: a three-profile scheme aligned with self-determination theory and a simple two-axis matrix defined by competence-signal sensitivity and transparency requirement. Practical implications include stating optionality and scope up front, maintaining purely binary early feedback, introducing a single mi-cro-hint after repeated failure, mapping outcomes to syllabus items, and ac-knowledging partial structure with lightweight code-aware indicators.