Law, on a global scale, is increasingly operated through the execution of code rather than the decision-making of institutional legal actors. A growing body of research points to the effects of this transformation, showing how the use of legal algorithms and code can affect judicial outcomes from the receipt of social welfare to prison sentencing. The logic that legal algorithms encode, however, does not emerge spontaneously. For legal text to be translated into computer code, the programmer must mobilize relevant domain expertise, interpret legal text, and negotiate team dynamics and interests. Each of these act as a mechanism through which the creators of legal algorithms shape this “black box.” Our study aims to build on computational law literature by following not only the effects, but the development of legal algorithms. How do legal and programming experts translate legal text into computer code? How does the socially and institutionally embedded nature of their work shape the tools they produce? We discuss preliminary results from an ongoing experimental study recruiting individuals and teams of experts in computer science and law to each develop a tool that interprets a legal statute and advises layperson users about the statute’s applicability to their case. We find considerable variation in how legal experts and programmers understand and encode the law, the resources they draw upon, and their confidence in implementing legal tools.