Less than 24 hours after Anthropic introduced its flagship model Claude Fable 5, AI commentator Riley Brown posted a demo that quickly spread across X. In the video, Brown said he used just two prompts to build a clone of Lovable’s app and showed it outperforming the official version in parts of the workflow. At the time of publication, the post had drawn more than 645,000 views, over 4,300 likes, and thousands of reposts and bookmarks.
A side-by-side test against Lovable’s official iOS app
In the nearly four-minute screen recording, Brown first asked the model to replicate Lovable’s overall visual interface and core programming logic. He then gave a second prompt instructing the cloned system to build a dark note-taking app similar in style to Notion. To compare the results, he entered the same app-building request into both the official Lovable iOS app and his own cloned version.
According to the demo, the Fable 5-powered clone delivered faster generation and smoother live preview performance. Brown also showed richer text-editing support, including headings, underlines, and tables. The cloned app could open a full preview in an external browser as well, and Brown argued that its UI detail had moved past the original app.
Five prompts, broader platform features
Brown later posted updates saying the project expanded sharply once he increased the instruction count to five prompts. He described the latest version as a more complete platform capable of building, editing, and live-previewing both web apps and mobile apps. He also said it includes a cloud sandbox that starts automatically when a project opens and shuts down after 60 minutes of inactivity.
His follow-up description added database integration, user authentication, and voice-to-text support. In another Swift-focused demo, Brown said the platform could generate native Swift source code directly from a request, then render a near-identical Apple Notes-style app inside the interface within seconds, while also offering a test installation link. He had previously claimed that a single prompt was enough to build a Replit-like mobile development tool.
AGI claim splits the developer community
Brown framed the moment in dramatic terms, calling Mythos “AGI.” That line became one of the most contested parts of the story. Skeptics argued that the demo still showed a system reproducing a relatively simple front end and note-taking workflow on top of the same large-model foundation, not a cross-domain autonomous intelligence system. In that view, the AGI label says more about hype than technical reality.
Others in the AI and Web3 developer crowd focused on the change in software output speed instead. Their takeaway was narrower and more concrete: tasks that once demanded a team and a long build cycle were being compressed into a handful of prompts and a short demo loop. That is why the Lovable clone clip gained traction so quickly. It turned a model launch into a live argument over whether advanced AI tools are starting to absorb the value proposition of no-code development platforms.