Claude now creates interactive charts, diagrams and visualizations

Original Article ↗ Hacker News Discussion ↗

Overall Reaction to New Visualizations

  • Many find the feature “magical”: Claude can now produce attractive, interactive charts and diagrams with minimal prompting, often unprompted when data is involved.
  • Users say the visuals significantly speed up understanding of complex data, processes, and portfolios, and are more polished than manual efforts or text-only outputs.
  • Others note Claude has long been able to emit HTML/JS or Mermaid code to create diagrams; this feels like a more integrated, UX-polished version of that.

UX and Account/Platform Frictions

  • Strong criticism of Anthropic’s UX:
    • No “Sign in with Apple” on the web despite supporting it on iOS, forcing awkward use of Apple’s relay emails on desktop.
    • Complaints about basic issues: code-block/backtick handling, mobile paste behavior, payment info removal, and inconsistent support for the new visualization API across web vs iOS.
  • Debate over responsibility:
    • Some blame Anthropic for poor UX and lack of Apple login on web.
    • Others argue users chose hidden emails and that Apple’s policies and relay system inherently create friction.

Product Quality vs Shipping Speed

  • Multiple comments describe Claude’s ecosystem as fast-moving but “janky,” with bugs, timeouts, and unreliable Claude Code remote/cloud sessions.
  • Some see this as a conscious “speed over robustness” tradeoff and consider it acceptable or even correct; others worry everything is being shipped half-baked.
  • Reports of very high engineer compensation and emphasis on rapid coding during interviews, plus heavy internal use of Claude to write code, are mentioned as context.

Correctness, Trust, and “Confident Slop”

  • Several worry that slick, interactive visuals will increase misplaced trust in LLM outputs, since visuals traditionally signal human effort and vetting.
  • Others argue diagrams and charts actually help users spot mistakes and reason about processes, especially when checking dense SOPs or complex systems.
  • Broader concern: LLM-generated visualizations can still fabricate numbers or logic, so they should ideally come with data and computation “receipts.”

Comparisons to Other Tools

  • Some say ChatGPT’s charting is clunky and requires very prescriptive prompts; Claude and Gemini are reported as better at interactive canvases.
  • A minority dismiss the excitement, noting spreadsheets and BI tools have done interactive charts for decades; the novelty is the natural-language interface, not the graphs themselves.