The tyranny of software is almost over. Since the first programmers wrote the first programs, users have been forced to live in worlds those programs created. Features are fixed, design is unchangeable, and wanting something better meant learning to code. That era is ending.
In late 2025, an update to Anthropic's Claude model turned Claude Code from a surprising code generator into one that rarely failed. Suddenly, all anyone needed was $20 a month and a half-formed idea to build functional software. Andrej Karpathy, an educator and OpenAI founding team member, called this new behavior "vibe coding." The vibes soared.
The Rise of AI Coding Tools
AI coding tools like Claude Code, OpenAI's Codex, GitHub Copilot, Cursor, Lovable, and Replit are changing how developers work. More importantly, they're enabling an entirely new kind of software: personal software built for oneself, not for venture capital or acquisition. This era is transforming our relationship with technology.
Personal software lets us create apps the way we used to make lists and spreadsheets. Managing a family budget? Build a custom app with every feature needed and none you don't. Can't stick with a to-do list app? Roll your own. Need a meal planner with grocery assignment for a family trip? Whip one up. No subscription fees, no marketing emails. It's your software.
Author and technologist Robin Sloan wrote in 2020 about his family messaging app: "There will be no sudden redesign, no flood of ads, no pivot to chase a userbase." Five years later, in late 2025, he updated: "I have changed literally nothing in the app, and it's glorious." Sloan now uses AI to make more home-cooked software, like a Ruby script that pulls data from Shopify and USPS to generate shipping labels for his olive oil company. It's hacky—only he knows how to run it—but it works perfectly for him.
Limits and Realities
Personal software has limits. Bespoke apps lack support lines, thorough testing, and security guarantees. Large companies won't ditch enterprise software for vibe-coded solutions. Most apps we download are fine. But everyone has edge cases—reasonable ways they'd morph software to exact needs. The problem is that everyone else has needs too, and none are identical.
The author's own edge cases are productivity tools. After cycling through GTD, CARE, PARA, and countless apps, they built a list of ideal features. No app checks every box. Developers explain: everyone has a unique list, and building for all makes a mess. As Doist CEO Amir Salihefendic noted, "It's ridiculously easy to build features right now. But if you just do it naively, you end up with a system that nobody can figure out." In the era of personal software, you don't have to build for everybody.
During the 2025 holidays, the author got a Claude Code subscription and built Timetable—an app combining calendar, notes, and tasks. The prototype took 20 minutes. Days of debugging followed: copying error codes, figuring out GitHub secrets, wiring up Supabase and Vercel. Then weeks to build a native mobile app. Design was a struggle: Claude Code loved purple gradients and hamburger menus. When asked to design an icon combining journal and planner, it produced an image that looked like a butthole. The next version was three horizontal lines.
Brian Lovin, a designer and engineer at Notion, says AI agents often suck at interfaces. He's learned to trust AI for adding settings tabs but not for initial scaffolding. Others trust it more. The Information reported App Store new apps grew 30% in 2025 after a decade of decline. By end of 2026, vibe coders could double the count to nearly 4 million. GitHub had its fastest growth year, with 80% of new users using Copilot within a week. Claude Code creator Boris Cherny knew it was a hit when the sales team used it.
Examples of Personal Software
Countless personal apps exist. Brenden made a command-line ranker for fantasy baseball. Nathan wrote a script to introduce renewable energy to Transport Tycoon Deluxe, a 1990s game. Anthony optimized Secret Santa assignments. Tucker mapped dog poop locations in his backyard. Allan tracked migraines. Brett tracked which of 102 stairs a package lands on. Most have a market of exactly one and revenue of zero.
The author built several apps: Timetable abandoned when bloated; Spring forgotten; Basket for collecting links and notes until Twilio bills came due. Each taught something. The key is that personal software doesn't have to be built from scratch. Knowledgeable developers are home cooks; the rest are Chipotle customers—choosing ingredients, not making the food. The future is building with models, not from scratch.
Infrastructure and Taste
Professional developers may focus on infrastructure—syncing, security, databases. Maggie Appleton, a designer at GitHub Next, coined "barefoot developers" for those who learn skills to help their communities. She advocates for open-source primitives—security, design, login, payment—that anyone can plug and play. Notion is a current example: users customize pages via AI, which writes macros rather than software from scratch.
This approach has risks. Your ideas might be bad. Users excel at identifying problems but struggle with solutions. When bespoke AI breaks things, who's at fault? Developers must balance customization with coherence. Craft CEO Balint Orosz says, "I think there's a responsibility to ensure we provide a coherent user interface, so if you like the core product, it feels like home." The goal is to let users say "make this bigger" and have it happen sensibly.
Much of current AI tech aims to make adaptation easy and universal. The Model Context Protocol exposes data to agents. More apps integrate with Claude and ChatGPT. The best answer is to build yourself a brand-new way to email. In this world, the most important asset is taste—knowing what you like. Music producer Rick Rubin said his success came from "the confidence I have in my taste, and my ability to express what I feel." Without that, you fall into "doom loops" of telling the chatbot only what you don't like.
The author has no opinions about databases but cares about typefaces and background colors. So the first useful vibe-coded app was a dashboard smashing together Raindrop, Todoist, Obsidian, and Google Calendar. Four API keys and an afternoon later, it worked—with plenty of "why doesn't that button do anything" and "let's try a color other than purple." The app looks like a paper planner. It will never be in the App Store. That's the beauty: you don't have to explain it. You don't have to accept software prescribed to you. If you know what you need and like, you can make things work exactly as you want. No learning to code required.
Source: The Verge News