Show HN: Scripton – Python IDE with built-in realtime visualizations
Visual Design & Core Functionality
- Many commenters praise the aesthetics and smooth visualizations, especially for numerical/data work.
- Real-time plotting from Python via a custom library that converts common data structures to an intermediate format and then to Observable Plot/JS is seen as compelling.
- Some ask for richer variable exploration and table-based inspection of dataframes/matrices, similar to MATLAB, RStudio, and Spyder.
Implementation & Architecture
- UI is TypeScript-based; initial React experiment was dropped in favor of custom VDOM and ultimately a lightweight, framework-free setup similar to VS Code.
- Editor uses Monaco but the IDE is not a VS Code fork. Author cites VS Code’s architectural constraints (extensions, IPC limits) as the reason for a from-scratch IDE.
- A C++ multithreaded node module handles low-latency IPC for large binary payloads (numpy arrays, images, tensors).
Extensibility, Widgets, and Tables
- There is a custom virtualized table capable of million+ row handling, currently backing the REPL; a full data table UI is planned but not yet exposed.
- Current widget set: text outputs, buttons, sliders. More controls (text inputs, checkboxes, table inputs) and richer UI-building are planned.
- Not user-extendable yet; some IPython rich outputs work, but ipywidgets are not supported.
Performance & Tensor Handling
- Zero-copy visualization of PyTorch tensors is not supported yet; prior experiments using IOSurfaces had limitations.
- IPC is claimed to minimize copies and handle large numerical arrays efficiently.
AI & Integrations
- No AI assistant at launch; GitHub Copilot integration is in active development.
- Some suggest using OpenRouter/Roo or positioning similarly to Cursor, but others note this tool is fully local, unlike cloud-LLM-focused editors.
Pricing, Licensing & Business Model
- $20/month Mac-only subscription is the dominant controversy.
- Multiple commenters reject subscriptions for local developer tools, preferring:
- One-time purchases or annual upgrade licenses.
- JetBrains-style “perpetual fallback” licenses (which the author says they’d like to add later).
- Concerns include:
- “Renting” tools vs owning them.
- Limited resources of a solo dev, risk of abandonment, and closed-source lock-in.
- Comparisons to PyCharm, Spyder, VS Code + plugins, and free tools like rerun or DearPyGui, which reduce perceived value at this price.
- Others argue $20/month is reasonable if it saves professional users even modest time, though several note “productivity ROI” is hard to measure and widely abused as justification.
Target Users & Competition
- Some say this is better aimed at “core engineers” and MATLAB users than general HN/SWE crowd.
- Several compare it to Spyder, Jupyter, Clerk (Clojure), RStudio, Data Formulator, and rerun; rerun in particular is highlighted as a strong open-source alternative for streaming, high-rate, multidimensional data visualization.
Platform Support & Adoption Risk
- Mac-only support is a major blocker; many potential users are on Linux or Windows, especially in traditional engineering and research environments.
- Some advocate building it as a standalone visualization/runtime tool or IDE-agnostic runner instead of a full IDE, to avoid competing head-on with VS Code/JetBrains and to integrate into existing workflows.
Usability & UX Details
- There is a variables panel that shows active symbols, shapes, and dtypes; deeper, tabular drill-down is “planned.”
- Hold-to-quit (Cmd+Q) behavior is divisive; author agrees to make it configurable.
- Several commenters express enthusiasm and say they’ll “keep an eye” on the project, contingent on pricing changes, platform support, and feature maturity.