AI-Powered Accessibility: Engineer Overcomes Vision Loss with Custom Chrome Extensions Built Using Claude Code

Joe McCormick, a principal software engineer at Babylist, faced a life-altering challenge when he lost most of his central vision due to a rare genetic disorder just before starting college. This unexpected obstacle led him to pivot from mechanical engineering to computer science, a decision that would ultimately place him at the forefront of AI-powered accessibility solutions. Joe now leads AI enablement at Babylist and is demonstrating how individual developers can use AI to create micro Chrome extensions that address specific accessibility needs that larger software often misses.

McCormick's approach involves utilizing tools like Claude Code, VS Code, and other AI assistants (including Gemini and ChatGPT) to rapidly prototype and deploy custom Chrome extensions. These aren't just theoretical exercises; they are practical tools that enhance his daily work and life. He demonstrated building extensions for image descriptions within Slack, AI-powered spell checkers, and link summarization—all optimized for screen reader accessibility and keyboard navigation.

The core of McCormick's strategy lies in harnessing the power of AI to automate repetitive development tasks. By leveraging "Claude Skills," he can extract common patterns and streamline the extension-building process. This accelerates development time and lowers the barrier to entry for others looking to create their own accessibility tools. He emphasizes the importance of removing "cognitive friction" by identifying and codifying repeatable patterns, making the development process more intuitive and efficient. The workflow includes the creation of a development plan, code review, modification, and the loading, testing, and debugging of extensions within Chrome's developer mode.

The implications of McCormick's work extend beyond personal productivity. It highlights the growing viability of personal software as a means to address highly specific accessibility needs. Mainstream software often strives for broad appeal, which can lead to compromises that leave some users behind. AI-powered tools like Claude Code empower individuals to fill these gaps with customized solutions. This approach is especially relevant in the context of rapidly evolving technologies like multimodal AI, which holds immense potential for transforming accessibility for visually impaired users.

McCormick's journey and demonstrated workflow serve as an inspiration and a blueprint for other developers looking to contribute to a more inclusive digital world. By embracing AI as a development partner, it is possible to create powerful, personalized accessibility solutions that significantly improve the user experience for individuals with disabilities. His advocacy for the use of ARIA (Accessible Rich Internet Applications), screen readers, and tools like Windows Subsystem for Linux, provide a comprehensive approach to ensuring digital content is usable by everyone.

Several resources were shared, including links to Chrome Extensions Documentation, Meta Ray-Ban Smart Glasses (as potential assistive devices), and further reading on Claude Skills. These resources combined with McCormick's example pave a practical path for others to follow. The potential for AI-driven accessibility solutions is vast, and McCormick's work serves as a compelling example of how individual ingenuity, combined with powerful AI tools, can create a more accessible and inclusive digital landscape.