[BUG] Claude can't output text before a widget OR place text between two widgets (really really really annoying)

Status Open
Maintainer reply None cached
Activity 0 comments · opened Jul 22, 2026

Preflight Checklist

  • [x] I have searched existing issues and this hasn't been reported yet
  • [x] This is a single bug report (please file separate reports for different bugs)
  • [x] I am using the latest version of Claude Code

What's Wrong?

Claude can't render text BEFORE a widget. It glitches.
Claude also can't interleave text BETWEEN two widgets.

For example, if I prompt this:

  1. Make a graph for age of u.s. presidents over time
  2. Then provide a short TEXT breakdown of the answer (outside the widget)
  3. Make another graph for height of u.s. presidents over time
  4. Then provide a short TEXT breakdown of the answer (outside the widget)

I get this as my output with a huge gap in between.

<img width="444" height="792" alt="Image" src="https://github.com/user-attachments/assets/1cd7ba61-21e7-480b-a3d9-2bc18d8fd4f2" />

What Should Happen?

It should be able to interleave text between.

This is really annoying because it means I can't build more detailed review projects out which would be awesome.

Error Messages/Logs

That's a rendering drop, not a template miss — which changes my last answer. I did generate the president text. But I interleaved them between two widget calls, and your client dropped every text block except the one after the final widget. Everything before and between the widgets vanished; only "The short version" survived.

So the failure mode is: prose interleaved with multiple widgets is fragile. The words that matter most — the extraction and the graded summaries — sat in the fragile zone.

Steps to Reproduce

  1. Make a graph for age of u.s. presidents over time
  2. Then provide a short TEXT breakdown of the answer (outside the widget)
  3. Make another graph for height of u.s. presidents over time
  4. Then provide a short TEXT breakdown of the answer (outside the widget)

Fails badly.

Claude Model

Other

Is this a regression?

No, this never worked

Last Working Version

_No response_

Claude Code Version

Claude Fable 5, claude.ai web interface

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

If you're filing the bug report, that's the detail to include: Claude Fable 5, claude.ai web interface, prose before an inline visualization call fails to render.

View original on GitHub ↗