This paper defines the concept of "improvement infrastructure" and "educational capital" for education, and it uses the case of the National Writing Project to develop an extended, data-based illustration of the design and generativeness of an improvement infrastructure. Since 1983 there have been multiple "waves" of educational reform that have sought to address fundamental problems in the system. In spite of many good efforts, there is still pervasive public dissatisfaction with the progress that has been made. Why has the nation not been able to do a better job of improving the education offered to US students? This paper asserts that educational improvement has until now been conceptualized and funded as expenditure, not as an investment. Direct services have been seen as a greater priority than the development and accumulation of various forms of capital that are needed to support the work of improvement. The paper argues that what is needed is a mind-shift away from expenditures that focus on direct services, and toward a greater focus on investing in the development of improvement infrastructures that generate educational improvement capital. Unlike short-term projects that address a singular issue and build little sustainable capacity, the NWP comes close to meeting the criteria for a robust, cost-efficient infrastructure for educational improvement. The NWP is ongoing, with the intent of developing and supporting a more or less permanent national network of local writing project sites that can provide a wide range of ongoing supports for teachers, addressing many different national and local challenges and, year by year, continuing to increase its own capacity to support teachers across the country. Appended are: (1) The Scale Of Work Of The National Writing Project; (2) Program Data By Age Of Site; and (3) NWP Service To Schools and Districts (Contains 18 figures, and 48 footnotes.)