This paper describes a project to develop a network-centric RTOS from scratch using formal methods. The (initial) purposes of the project was to get acquainted with the use of formal methods for software engineering and to obtain a trustworthy RTOS as a component for building networked embedded systems. The work was done by a small, distributed team that had no prior experience on using formal methods and with a small budget. The outcome is that the use of formal methods is most useful as an architectural design method, perhaps more than as a formal verification of software code. The resulting software has many properties that were not anticipated at the beginning and would likely not have been achieved without the use of Formal Methods.