How this visually impaired engineer uses Claude Code to make his life more accessible | Joe McCormick
How I AI
Feb 16
How this visually impaired engineer uses Claude Code to make his life more accessible | Joe McCormick
How this visually impaired engineer uses Claude Code to make his life more accessible | Joe McCormick
How I AI
Feb 16
This episode features Joe McCormick, a principal software engineer at Babylist who navigated a significant vision loss before college and now leads AI enablement—building custom, accessible tools that empower his daily work and life.
Joe shares how he uses AI—not for full automation, but as real-time, practical assistance—to interpret images, summarize links, correct spelling, and describe visual content in Slack. He demonstrates building lightweight, screen-reader-accessible Chrome extensions in under 25 minutes using Claude Code, emphasizing keyboard shortcuts, ARIA-compliant UIs, and iterative debugging. His approach centers on 'personal software': tailored solutions that address specific accessibility gaps mainstream tools overlook. By leveraging Claude Skills, VS Code Copilot audio, and prompt engineering guided by context—not just instructions—he lowers the barrier to creating robust, accessible tools. Joe highlights how standardizing patterns reduces cognitive friction, especially for screen reader users, and stresses fluency through focused practice: mastering one domain while learning AI tools yields outsized returns. He also notes emerging multimodal capabilities—like Gemini’s live-share for reading physical books—that deliver equitable, real-world accessibility beyond hardware alternatives. Ultimately, Joe champions AI as a force multiplier for independence, creativity, and inclusion in software engineering.
00:00
00:00
Joe uses Gemini and ChatGPT to handle images and answer follow-up questions
02:34
02:34
With the rise of AI, the gap between sighted and visually-impaired software engineers is closing
04:50
04:50
AI allows people to build customized accessibility solutions for themselves
06:10
06:10
Image description tool in Slack enables independent visual understanding without relying on others or sight
10:45
10:45
Joe built an AI spell checker that works in any web input field, sending text to OpenAI to correct typos
13:12
13:12
Keyboard shortcuts enable efficient micro app triggering in AI tools
17:15
17:15
Engineers can use AI to write good PRDs without custom prompts, relying on foundational models
20:28
20:28
Claude Code now supports editing prompts in a text editor for better screen reader accessibility
25:30
25:30
Ctrl G shortcut enables fast plan editing in the editor
27:50
27:50
Claude Code standardizes inputs, reducing cognitive friction for users
31:42
31:42
Building personal software to automate recurring tasks shrinks the payback period to virtually zero
34:56
34:56
Chrome developer mode allows installing extensions from a local computer for easy testing
39:09
39:09
Moving JSON out of the prompt ensures the response object is valid JSON for reliable Chrome extension parsing
40:44
40:44
The extension parses shared Slack links to extract key takeaways using AI-generated code with minor human tweaks.
44:56
44:56
Gemini's live-share feature allows a visually impaired parent to read any physical book with his kids
