Detailed Analysis
A law student's viral anecdote about turning an unwieldy nail polish spreadsheet into two functioning mobile apps illustrates a use case that has quietly become one of the most common ways non-engineers interact with Claude: converting personal data hoards into custom software. The user had spent years building an elaborate Google Sheet—20 tabs deep, cataloguing over 1,000 polishes along with stamping plates and manicure usage history, complete with a hand-built search function. The spreadsheet worked on a laptop but was nearly unusable on mobile, where the Google Sheets app struggled with storage and search performance. Rather than pay for an existing inventory app or manually re-enter a thousand-plus records into a third-party tool, the user asked Claude to brainstorm solutions, and Claude proposed building a custom HTML mobile app that syncs directly with the existing spreadsheet.
The result—two purpose-built apps, one for polish and one for stamping plates, complete with a randomizer feature and statistics dashboards—did not emerge from a single prompt. The author is explicit that it took "a lot of trial and error and a lot of different iterations and troubleshooting" to get there. This detail matters because it captures an accurate picture of what "vibe coding" with an AI model actually looks like in practice: not a magic one-shot generation, but an iterative back-and-forth in which a non-programmer describes desired behavior, tests the output, reports what's broken or missing, and refines the request until the tool matches their mental model of what they wanted. The success also depended on the underlying data being meticulously structured already—every field manually typed, every formula in place—which gave Claude clean, well-organized source material to build against rather than messy, inconsistent data that would have required significant cleanup first.
This story is emblematic of a broader shift in who benefits from AI coding capabilities. Much of the public discussion around Claude and coding centers on professional software engineers using it to accelerate work in large codebases, refactor legacy systems, or ship production features faster. But cases like this highlight a parallel and arguably larger population: hobbyists, students, and professionals in unrelated fields (here, someone in their final year of law school) who have long had a "good enough" spreadsheet or manual system and lacked the technical skill—or budget—to turn it into real software. Anthropic and other AI labs have increasingly marketed exactly this capability, positioning natural-language coding assistants as tools that let anyone with a clear idea and some patience build functional apps without hiring a developer or learning to code from scratch.
The appeal of building a bespoke tool rather than adopting an existing app is also notable. The user explicitly rejected commercial nail polish inventory apps because none matched her exact workflow, and because migrating over a thousand entries risked data loss or formatting errors during import. Custom-built software that syncs to data the user already owns and trusts sidesteps that migration risk entirely and produces a tool tailored precisely to idiosyncratic needs—a randomizer for outfit-style inspiration, a stats page, cross-referencing polish with plates and past manicures. This kind of long-tail, personally significant but commercially unviable software is exactly the category traditional app development economics have never served well, since no company would build a nail-polish-and-stamping-plate tracker for a niche hobbyist market. AI coding assistants collapse that economic barrier, letting individuals generate one-off, highly personalized tools for their own passion projects, which is likely to keep surfacing in exactly this kind of grassroots, enthusiastic user testimonial rather than in enterprise case studies.
Read original article →