A GitHub repo for sketching app screens has 6,500+ stars — but does anyone actually use it?
M3E Canvas promises to turn rough app sketches into AI-coding prompts, and it's racked up stars at a startling rate. Here's what's actually confirmed and what isn't.
What it is
M3E Canvas is a free, open-source tool that lets you sketch app screens in the style of Google’s Material 3 Expressive design language directly in your browser, link those screens together so you can tap through a rough prototype, then export the whole thing as a text prompt. That prompt is meant to be handed to an AI coding assistant such as Claude Code, Codex, Gemini CLI or Cursor, which then does the actual work of building the app.
It’s a TypeScript project, released under the MIT licence, with the usual scaffolding of a serious open-source repo: a contributing guide, a code of conduct, a security policy, and documentation translated into Japanese, Chinese and Korean. There are also files aimed squarely at AI agents — an AGENTS.md and a CLAUDE.md — suggesting the tool was built with an eye on being consumed by bots as much as humans.
The number that’s turning heads
What’s driving attention isn’t the software itself so much as the speed of its uptake. The repository has picked up 6,579 stars and 666 forks in under a fortnight — a pace that, on GitHub, usually gets read as “developers are genuinely excited about this.”
That’s worth taking with a pinch of salt. Stars are a popularity signal, not a quality one. They cost nothing to give, can be encouraged by prominent sharing on social media or newsletters, and in some corners of GitHub have been artificially inflated by bot farms or reciprocal-starring schemes. None of that is alleged here specifically — but a headline star count, on its own, tells you that a link has spread, not that a tool works well, is maintained long-term, or does what its README promises.
So who is actually behind it, and does it work?
The GitHub page doesn’t offer much detail on the team or company behind the “lnkiai” account, beyond the code itself and its licensing paperwork. There’s no independent review, benchmark or user testimonial in the source material to confirm the tool reliably produces usable prompts, nor how well AI coding assistants actually perform when fed its output. The demo shown on the repo — sketching a recipe app, re-theming it, and running the AI-generated result on Android — is a proof of concept from the maintainers themselves, not a third-party test.
In short: the code exists, it’s real, and it’s popular by GitHub’s own metrics. Whether it’s genuinely useful for turning a rough idea into working software, versus just a slick demo that surfed the current wave of “vibe coding” hype, isn’t something the star count can tell you.
What it means for you
If you’re a developer or designer curious about prototyping with AI tools, M3E Canvas is free to try and costs nothing but your time to test. If you’re just seeing the star count shared around and assuming it must be good because everyone else seems to think so, be a bit more careful — GitHub popularity is a starting point for curiosity, not a verdict.
The sensible move is the boring one: clone it, try it on a small project, and judge for yourself whether the prompts it generates actually save you work. Six thousand stars buys attention. It doesn’t buy proof.