More than 10% of the population has dyslexia, and most are diagnosed only after they fail in school. My work is changing this through scalable early detection and tools that help people with dyslexia read and write better. To detect dyslexia, I am developing machine learning models that predict reading and writing difficulties by watching how people interact with my web-based game Dytective. My experiments have revealed differences in how people with dyslexia read and write, and I have developed a series of tools that integrate these results to help people with dyslexia read and write better. These tools are used by tens of thousands of people, which apart from supporting users, also serve as living laboratories in which to develop and prove techniques for detection and intervention. Moving forward, we are working with schools to put our approach into practice at scale to finally eliminate school failure as a primary way dyslexia is diagnosed.