The Vibe Coding Era: Why We Need RSS to Share Personal Micro-Apps

When Programming Becomes as Simple as Writing a Blog Post

Matt Webb recently raised a thought-provoking point: as vibe coding dramatically accelerates app development, we need an RSS-like mechanism to share the constantly emerging personal tools and micro-apps.

Vibe coding refers to the development approach of rapidly generating applications by describing requirements in natural language with the help of AI programming assistants. This concept was coined by Andrej Karpathy (former Tesla AI Director, OpenAI co-founder), a prominent figure in the AI field, in early 2025. He described an entirely new programming experience: instead of writing code line by line, developers describe their desired functionality in natural language, and AI coding tools like Cursor, GitHub Copilot, and Windsurf automatically generate complete code implementations. The developer's role shifts from "code writer" to "intent communicator" and "result reviewer." Karpathy himself even stated that in this mode, he would "fully surrender to the vibes, embrace exponentials, and forget that the code even exists." This approach works thanks to the rapid progress of large language models (LLMs) in code generation capabilities—models like GPT-4 and Claude 3.5 can already understand complex programming contexts and generate structurally complete, ready-to-run application code. This is fundamentally changing the production cadence and distribution logic of software.

A Paradigm Shift in App Development: From "Shipping Products" to "Sharing Posts"

Matt Webb's core insight is this:

When vibe coding accelerates app development, apps become more personal, more contextual, more frequent. Publishing a tool or micro-app is no longer like launching a website—it's more like writing a blog post.

This analogy is remarkably precise. Traditional software development involves lengthy cycles—requirements analysis, design, development, testing, deployment, operations. But when AI can help you generate a fully functional small tool in minutes, the "weight" of software drops dramatically, transforming it into a lightweight form of expression.

The Distribution Problem for Micro-Apps Surfaces

When individual developers are producing new micro-apps every week or even every day, a critical question emerges: How are these tools discovered, shared, and installed?

Matt Webb envisions a scenario where every developer's tools page offers an RSS feed, with each entry featuring an "install" button. You could subscribe to a developer's tool stream just like subscribing to a blog, discovering and using their new creations as they appear.

RSS (Really Simple Syndication), born in 1999, is one of the most important open protocols from the early internet era. It allows websites to publish content summaries in a standardized XML format, enabling users to aggregate subscriptions from multiple sources through RSS readers (such as Google Reader, Feedly, NetNewsWire) without visiting each site individually. During the golden age of blogging (2004–2010), RSS formed the infrastructure for decentralized content distribution, allowing independent bloggers to reach readers without relying on any platform. Although the rise of social media marginalized RSS for a time, it has been experiencing a renaissance in recent years as people grow tired of algorithmic recommendations and value information autonomy. Matt Webb's proposal to apply this mechanism to micro-app distribution is essentially a call for a decentralized, user-controlled approach to software discovery.

Simon Willison's RSS Implementation: From Idea to Reality

Inspired by Matt Webb, prominent developer Simon Willison immediately put the idea into action. He had Claude (Anthropic's AI assistant) add an Atom feed and corresponding icon to his /elsewhere/tools/ page.

Simon Willison is the co-creator of the Django web framework and the author of Datasette, a data exploration tool. In recent years, he has become one of the most active practitioners in the AI/LLM tooling space, developing the LLM command-line tool (which allows calling various large language models directly from the terminal) and numerous practical utilities built around AI. His tools.simonwillison.net website hosts dozens of single-page micro-apps written in HTML and JavaScript, covering text processing, data conversion, image analysis, and many other functions—many of which were completed in extremely short timeframes with the help of AI assistants like Claude. His workflow—rapid ideation, AI-assisted implementation, instant publishing, open-source sharing—is a textbook embodiment of the vibe coding philosophy.

This page aggregates the various small tools he publishes on tools.simonwillison.net. Now, anyone can subscribe to his tool update stream via an RSS reader—as naturally as subscribing to his blog posts.

It's worth noting that this Atom feed itself was completed with AI assistance (he submitted a GitHub PR documenting the entire process), which perfectly demonstrates the efficiency of vibe coding. Atom is the successor standard to RSS, published by the IETF in 2005 as RFC 4287. It resolved incompatibility issues between RSS's multiple versions and provides stricter specifications and richer metadata support, which is why many modern applications prefer it.

Unsolved Problems in the Micro-App Ecosystem

"Install to Where?"

Matt Webb himself acknowledges an unresolved question: after clicking the install button, where exactly does the app get installed?

This touches on a core gap in the current ecosystem. Our existing solutions each have limitations:

• Browser extension stores — but micro-apps aren't necessarily extensions
• PWA (Progressive Web Apps) — but the experience and discovery mechanisms are still not ideal
• Various SaaS platforms — but they're too heavyweight for micro-apps

PWA is a technical concept introduced by Google in 2015, designed to give web applications a near-native app experience. Its core technology stack includes Service Workers (enabling offline caching and background sync), Web App Manifest (defining app icons, splash screens, and other metadata), and HTTPS secure connections. Users can "add to home screen" to make PWAs look and behave like standalone apps. However, PWAs face multiple obstacles in practice: Apple's long-standing limited support for PWAs on iOS leads to inconsistent cross-platform experiences; PWAs lack a unified app-store-style discovery mechanism, making it difficult for users to browse and search as they would in the App Store; additionally, PWAs have a more restricted permissions model compared to native apps, unable to access certain system-level APIs. These limitations mean that while PWAs are technically viable, they remain insufficient as a unified runtime for micro-apps.

Perhaps what we truly need is an entirely new "personal app runtime"—a lightweight container that can install, run, and manage these vibe-coded micro-apps with a single click.

Trust and Security Challenges

When installing an app becomes as simple as clicking a link, security concerns are amplified accordingly. How do we ensure that these rapidly generated apps don't introduce security risks? This requires new sandboxing mechanisms and trust models to protect users.

Sandboxing is a core concept in computer security, referring to running untrusted code in an isolated, restricted environment to prevent it from accessing or damaging the host system's resources. Modern browsers already implement multiple layers of sandboxing—each Chrome tab runs in an independent process, and JavaScript code is confined within strict security boundaries. But when micro-apps need to access local files, network requests, user data, and other resources, sandbox boundaries need to be carefully designed. Current reference approaches in the industry include: WebAssembly (Wasm) providing near-native performance with memory safety guarantees; Deno runtime's "no permissions by default" model (code cannot access the file system, network, etc. by default—explicit authorization is required); and Docker containerization technology providing process-level isolation. For vibe-coded micro-apps, an ideal trust model would likely need to combine code signing (verifying developer identity), permission declarations (apps explicitly listing required permissions), community auditing (community review of open-source code), and runtime monitoring (anomalous behavior detection) to strike a balance between convenience and security.

The Bigger Picture: Software Is Transforming from Product to Content

This discussion reveals a deep transformation already underway: Software is shifting from industrial product to handcraft, from product to content.

Just as blogging enabled everyone to become a publisher, vibe coding is enabling everyone to become a software developer. And once the barriers on the production side disappear, distribution and discovery mechanisms become the next thing urgently in need of reinvention.

RSS once solved the content distribution problem for the blogging era; now we need similar open protocols to handle the explosion of micro-apps. This is perhaps one of the most important infrastructure questions of the AI programming era. History always rhymes—just as the Web 2.0 era gave birth to RSS, Atom, OpenID, and a series of open protocols to support a decentralized content ecosystem, the AI programming era likewise needs a new set of open standards to define micro-app description formats, permission models, distribution protocols, and runtime environments. Whoever defines these standards first may become the infrastructure builder of the next era.

Key Takeaways

• Vibe coding makes app development as frequent and lightweight as writing blog posts, requiring new distribution mechanisms
• Matt Webb proposes using RSS feeds to share personally developed micro-apps and tools
• Simon Willison, inspired by the idea, immediately used Claude to add an Atom feed to his tools page
• Core unsolved problem: where should these micro-apps be "installed"—a unified runtime environment is missing
• Software is transforming from a product form to a content form; distribution and discovery mechanisms need to be reinvented