Video summary: After a publicly leaked copy of Anthropic’s Claude Code, the creator downloads and inspects the repository, uses OpenAI’s Codeex to analyze the agent, then modifies its verbs and persona. They build a local fork named Riley Code, create a desktop Electron-like UI, and iterate on features — embedded browser, project/session handling, tool-call display, and persistent personality prompts — demonstrating hands-on reverse engineering and UX customization.
- Source inspection: Downloaded the leaked Claude Code package and used Codeex to summarize the agent, CLI entry points, multi-agent features, and runtime modes.
- Behavioral changes: Replaced action verbs, added a dedicated persona prompt file, and tuned the agent to respond in lowercase with a specific ’19-year-old intern’ tone.
- Local fork & runtime: Created a runnable fork called Riley Code, executed it locally with an Anthropic API key, and validated tool/skill invocations and session memory.
- UI/UX prototyping: Built a desktop app shell, adjusted tool-call visuals, added an embedded browser for links/files, workspace/project handling, and persistent chat metadata (soul.md).
Quotes:
Anthropic made a huge mistake — they accidentally leaked Claude Code.
Anthropic, please don’t hurt me.
I changed its personality to speak in all lowercase like a 19-year-old intern.
Statistics
| Upload date: | 2026-04-04 |
|---|---|
| Likes: | 127 |
| Comments: | 32 |
| Statistics updated: | 2026-04-04 |
Specification: Claude Code Was Leaked… So I Had to Test it
|
