An infinite canvas for code exploration

Original Article ↗ Hacker News Discussion ↗

An “infinite canvas” tool for exploring large C/C++ codebases as interconnected cards is drawing interest from developers who see it as a way to preserve mental context, map complex call flows, and share explanatory code maps with others. Many respondents, however, are wary of mandatory signup, unclear pricing and terms, UI issues like back-button hijacking, and the idea of sending private code to a hosted service. There is strong demand for local or self-hosted options, IDE (especially VS Code) integration, broader language support via LSP/Tree-sitter, and features closer to general-purpose whiteboards or prior tools such as Code Bubbles and reverse‑engineering UIs.

Overall reaction

  • Many commenters find the “infinite canvas” idea compelling and say they’ve been wanting something similar for years.
  • Several would pay for a good implementation, especially for private codebases.
  • Others are unconvinced it adds much beyond modern IDE navigation, and want clearer “superpowers” vs existing tools.

Concept & value proposition

  • Tool slices a repo into atomic units (functions, data structures) and shows an expandable graph of definitions and references.
  • Goal: preserve and externalize mental context when tracing deep call/control flows, and produce persistent code maps that explain systems to others.
  • Current focus: large C/C++ codebases using clang; Linux, LLVM, Godot are indexed.

Feature requests & integrations

  • Strong demand for:
    • VS Code (and other IDE) integration, ideally as an LSP/intellisense-powered plugin.
    • Local/standalone app or self-hosted mode.
    • Editing within the canvas, not just viewing.
    • Keyboard navigation, bookmarks, quick jumping between “cards.”
    • Notes, arbitrary links between snippets, and integration with other artifacts (logs, UI screenshots, issues).
  • Suggestions to support Unreal Engine, Jupyter-style workflows, and richer data-flow visualization (like Unreal Blueprints).

Language support & technical choices

  • Currently C/C++ via libclang.
  • Some suggest tree-sitter + LSP for broader language support; creator replies that clang’s semantic analysis and low-level control were important for initial goals.

UX and onboarding issues

  • Multiple complaints about:
    • Mandatory signup wall before trying the tool or seeing pricing/ToS.
    • Back-button “hijacking” and odd history behavior.
    • Video auto-play issues, missing controls, and it looking like a static image.
    • Layout issues (overlapping boxes, unclear placement), and desire for better color-coding and long-function handling.
    • Sign-in errors, JS exceptions, redirect loops, and mobile rendering problems.

Trust, privacy, and deployment model

  • Many are reluctant to send proprietary code to a new hosted service, especially in the current AI/data climate.
  • Several state they would strongly prefer local, offline, or open-source versions.
  • Creator clarifies they currently only index open-source code and want to prove scalability before handling private repos; ToS/privacy details for private code are flagged as essential.

Comparisons to existing tools & prior art

  • Mentioned analogues: IDE navigation (JetBrains, VS, VS Code), older tools like Sourcetrail and Code Bubbles/Debugger Canvas, reverse-engineering UIs, and generic whiteboard tools (Figma/Excalidraw/Obsidian canvases).
  • Some note that reverse-engineering tools have had similar graph views for years and wonder why this hasn’t spread to mainstream development.
  • One detailed comment describes an existing “manual” workflow (IDE + whiteboard) that Territory could partially automate, but notes whiteboards currently integrate better with non-code context.