AI’s Hallucinations Are Over

**As I understood you assume that our solution after generation of unit tests checks them and returns only those tests which did not fail. Not a bad solution, but we can’t make such a check. To perform generated unit tests sources of the target method are required, however we do not send user’s source code to our server - we stated this important security point in the provided instruction.** 

**I wanna direct your attention to the fact that the work of our solution is completely stable in the context of delivering a result for the received task: no matter how many times and when you run the generation of unit tests for a specific target method, the generated result (written unit tests) will always be the same.** 

\
This feature of our tech approach is absent in most solutions on the market of AI services, so they still can't be used as an industrial standard, just by single innovators to solve some problems.

\


**Firstly, there is a free version - __[link](https://plugins.jetbrains.com/plugin/22264-marvin-free)__. Secondly, we published it to demonstrate the absence of hallucinations. Did you find hallucinations? A full commercial version will be later. Thirdly, for methods that need coverage by two or three tests, it writes at the same level as competitive solutions, and for those methods that need four or more unit tests, it writes better than any competitor.**

Oleksandr Zabashnyi