We built you a new one: Scrimba-style CodeCasts inside a branching AI roadmap, so you learn by building instead of tabbing between YouTube and VS Code.
The difference
Scrimba gives the teaching loop. RoadRunners adds the map: pick a goal, choose the next build, pivot when interests change, and keep moving through one project.
Roadmap creator
State what you want to learn, build, or become.
Choose skills and topics that shape the first route.
AI proposes a small set of valid feature checkpoints.
Branching choices
Every option is a concrete project outcome.
Move from React to Swift, Python, or another track when it fits.
Deferred choices resurface, and old branches stay visible.
CodeCasts
One timeline syncs narration, captions, slides, files, and code.
The editor, file tree, tabs, terminal, and run controls stay in one surface.
Challenge gates stop playback so the learner has to type, run, or skip.
Saved CodeCasts keep timeline position, active file, and code.
Workspace
Code appears step by step, then becomes editable.
Run the project, inspect output, and preview web work.
Daytona-backed execution with Pyodide fallback when needed.
Verification
Run status, stdout, stderr, exit code, and timeout state.
Assertions, HTTP probes, browser checks, and expected output.
One bounded judgment only when objective checks need context.
Progress
The map shows current work, pivots, and archived branches.
Concepts are labeled introduced, practiced, or verified.
XP, levels, and streaks reward momentum without replacing proof.
What we used
App Router, TypeScript, Tailwind CSS, shadcn/ui.
Auth, Postgres, Storage, RLS, saved CodeCasts.
Cloud sandboxes, terminal execution, snapshots, previews.
Branching roadmap choices, guides, and advisory checks.
Create a roadmap, pick a feature, play a CodeCast, pause for the challenge, run your code, verify it, and unlock the next build.