AI Empowers Individuals to Build Personalized Software Solutions

The immediate power of AI lies not just in solving problems, but in empowering individuals to solve their own problems, often with surprising speed and efficiency. This conversation with Joe McCormick, a visually impaired engineer, reveals a profound shift: AI is democratizing the creation of highly personalized tools, particularly for accessibility, that were previously economically unviable or technically out of reach. The hidden consequence is the erosion of the "expert gap" in software development, where individuals can now bypass traditional barriers to build bespoke solutions. This is essential reading for engineers, product managers, and anyone interested in the practical, human-centered applications of AI, offering a tangible blueprint for turning personal friction into a competitive advantage through custom software. It highlights how AI can level the playing field, not just in terms of access, but in the very act of creation.

The Personal Software Revolution: Building Bridges with AI

The traditional software development landscape often overlooks the nuanced needs of individuals, particularly those facing accessibility challenges. Building custom solutions for niche problems can be economically unfeasible, leaving many users to adapt to tools that don't quite fit. Joe McCormick, a principal software engineer at Babylist, offers a compelling counter-narrative, demonstrating how AI, specifically Claude Code, is democratizing the creation of highly personalized tools. His journey, marked by vision loss, has led him to leverage AI not just to navigate the world, but to actively shape it through custom-built micro Chrome extensions. This isn't just about making life easier; it's about reclaiming agency and efficiency in ways previously unimaginable.

From Vision Loss to Code Creation: The AI Advantage

McCormick’s transition from mechanical engineering to computer science was catalyzed by vision loss. He discovered that coding offered a similar creative outlet, but the playing field was uneven. AI has dramatically shifted this dynamic.

"obviously as ai took off became even more um equivalent and the gap between um i think software engineer for sighted person and an individual impaired person is is closing -- day by day."

This closing gap is most evident in the realm of personal software. While broad AI-powered browsers exist, McCormick emphasizes the power of specialized, "drill-like" tools over a general "Swiss Army knife." He has built micro Chrome extensions that address specific pain points in his daily workflow, particularly within Slack, where he spends a significant amount of his non-coding time. These tools are not just about accessibility; they are about optimizing efficiency for him.

One such tool is an image description extension. For someone using a screen magnifier, parsing images can be tedious. McCormick’s extension, triggered by a keyboard shortcut, uses AI to describe images within Slack messages. This immediate feedback loop allows him to understand visual content without the manual effort of zooming or asking for explanations, saving valuable time and reducing friction. The ability to then ask follow-up questions, leveraging the same AI, further streamlines information retrieval.

"it's just a nice way for me to not have to eventually push back and work with people and get some answers to my questions -- as i go on this too."

Another practical application is an AI-powered spell checker that works across any web input field. While Grammarly and similar tools exist, McCormick’s custom solution is free, accessible, and integrated directly into his workflow. He emphasizes a hyper-focused prompt that ensures only typos are corrected, preserving the original content. This addresses a personal typing challenge, exacerbated by his current preference for decorative nails, transforming a potential source of frustration into a seamless part of his communication process. The efficiency gain here is palpable: a quick shortcut, an AI check, and the message is ready to send.

The "Under 25 Minutes" Workflow: Building a Link Summarizer

The true power of AI, as McCormick demonstrates, lies in its ability to drastically reduce the development time for bespoke tools. He walks through the creation of a Slack link summarization extension, a task that took him approximately 25 minutes from concept to working prototype. This rapid iteration cycle is a direct consequence of AI-assisted development.

The process begins with defining a Product Requirements Document (PRD). McCormick uses AI to generate this, dictating his needs into VS Code’s co-pilot integration. This initial PRD outlines the core functionality: a Chrome extension for Slack that, upon a keyboard shortcut, identifies links in a message, extracts their content, sends it to an AI for summarization, and presents key takeaways in a screen-reader-accessible modal.

"we want to build a simple prd here for a locally run chrome extension where the job is when in slack in chrome and hovering over a message that has focus you can run the keyboard shortcut control shift one and that will look for any external links in the message if any are found open them up in hidden tabs and extract that content and send it over to openai to summarize and extract three to five key takeaways from the article and return those to the user in screen reader accessible modal which includes the article's title and a link out in a new tab to view the article."

The next crucial step involves leveraging Claude Code, specifically a custom "Claude Skill" designed for building Chrome extensions. This skill encapsulates common patterns and boilerplate code, significantly accelerating the development process. McCormick highlights the utility of the Control + G shortcut within Claude Code, which allows him to edit prompts and plans directly in a text editor, a much more accessible workflow for screen reader users. He also notes the importance of managing API keys securely, using symbolic links to a shared configuration file across his extensions.

The development plan itself is reviewed and refined, again using the Control + G shortcut for accessibility and efficiency. McCormick emphasizes that for personal software, the focus shifts from production-level code quality to pure functionality: "What matters is does it work?"

The extension is then loaded into Chrome via developer mode. A crucial step involves refreshing the Slack tab to ensure the new extension is recognized. The initial test reveals a minor issue: the output is JSON, not a cleanly formatted summary. This is where the iterative nature of AI-assisted development shines. McCormick uses a custom screenshot slash command (necessitated by his Windows Subsystem for Linux environment) to capture the output and sends it back to Claude Code for refinement. The AI is instructed to adjust the response format to properly handle JSON, leading to the desired, accessible modal output.

The Payoff: Efficiency, Accessibility, and Control

The ability to build such tools rapidly has profound implications. McCormick notes that the return on investment for personal software has become "insane." A task that might take minutes to perform repeatedly can now be automated with a tool built in under an hour, yielding a payback period of mere days or weeks. This shifts the focus from simply doing a task to automating it, creating significant leverage.

"it saves me three minutes a day and it takes me 30 minutes to build -- the payback period has just become insane for a lot of this tooling."

Furthermore, the emphasis on accessibility in these AI-generated tools benefits everyone. By instructing AI to create screen-reader-accessible modals, developers can tap into well-documented accessibility standards (like ARIA roles) that foundational models are adept at implementing. This proactive approach to accessibility, embedded within the development process, ensures that tools are not only functional but inclusive.

The multimodal capabilities of AI are also opening new frontiers. McCormick shares a heartwarming example of using Gemini’s live-sharing feature to read books to his young children, overcoming his inability to read Braille. This demonstrates AI’s potential to bridge personal limitations and foster deeper connections.

Key Action Items

• Embrace Personal Software Development: Identify recurring points of friction or inefficiency in your daily workflow and consider building a small, AI-assisted tool to address them.
  • Immediate Action: List 2-3 recurring tasks that consume more than 5 minutes of your time daily.
• Leverage AI for PRD Generation: Use AI tools to draft Product Requirements Documents for your personal projects. Experiment with dictation and clear prompts.
  • Immediate Action: Practice dictating a simple PRD for a hypothetical tool.
• Explore Custom AI Skills/Workflows: If you find yourself repeating development patterns, investigate creating custom skills or reusable AI workflows (e.g., Claude Skills).
  • Within the next quarter: Research and experiment with creating a simple custom skill for a common task.
• Prioritize Accessibility in AI Outputs: When prompting AI for code or user interfaces, explicitly request screen-reader accessibility and adherence to standards like ARIA.
  • Immediate Action: When using AI for UI generation, add a prompt like "ensure the output is fully screen-reader accessible."
• Invest in Keyboard Shortcuts: For AI-powered tools and custom extensions, map keyboard shortcuts to trigger actions, significantly improving efficiency.
  • Within the next month: Identify one AI tool you use frequently and explore its keyboard shortcut options.
• Understand the "Payback Period" of Automation: Recognize that the time invested in building an automation tool often yields a far greater return than the time saved by performing the task manually over time.
  • This pays off in 12-18 months: Shift your mindset to prioritize learning automation skills over performing repetitive tasks.
• Experiment with Multimodal AI: Explore AI tools that integrate different modalities (text, image, voice) for novel applications, especially in areas like content creation or information access.
  • Within the next quarter: Test a multimodal AI tool for a task outside your usual workflow.