Bug report 01
Shared Chat Artifact Preview Can Crash the Share
The problem
A shareable ChatGPT conversation can become unsafe to open when a downstream presentation layer previews, renders, or source-boxes a massive generated HTML artifact. The assistant may only intend to provide a download link, but the client UI can transform that into heavy inline content.
Why it matters
- The user wanted to share a conversation, not mass-distribute a browser crash.
- Single-file apps may contain hundreds of thousands of lines of code, comments, docs, CSS, and runtime logic.
- The assistant cannot reliably control downstream preview behavior with wording alone.
Requested product fix
| Need | Behavior |
|---|---|
| Download-only artifact mode | Never render, preview, parse, or source-box the file in chat. |
| Large HTML safety threshold | Automatically ZIP or block inline preview for huge HTML/source artifacts. |
| Shared-chat safety | Shared conversations should remain lightweight and openable. |
2000-character report
ChatGPT shared-chat/artifact presentation can make a conversation unsafe to share. In a production app workflow, the assistant provided a direct link to a massive single-file HTML app. Downstream ChatGPT presentation layers attempted to preview/render/source-box the artifact, making the chat heavy/crashy. The assistant cannot reliably know or control which clients will preview large artifacts. Users need a safe “download-only / never preview / never render” artifact mode, especially for single-file apps and generated HTML. Please add product-level protections: disable preview for large HTML/code artifacts, honor ZIP-only handoff, warn before rendering huge files, and keep shared chats safe to open.