mindscript 0.2.14

An experimental programming language combining formal and informal computation.

An experimental programming language combining formal and informal computation.

© 2024, 2025 Daios Technologies Limited

Documentation: https://mindscript.daios.ai
Source code: https://github.com/DAIOS-AI/mindscript
Playground: https://www.daios.ai/playground

Description

MindScript is a programming language that seamlessly integrates both formal and informal computation.

MindScript lets programmers code directly when the method for accomplishing a task is clear. Conversely, when they know what they want but not how to achieve it, developers can simply describe their intent. Indeed, for certain functions, such as analyzing the sentiment of a sentence, there might not exist a concrete implementation at all. The syntax is designed to make such specifications straightforward.

A distinctive feature of MindScript is its dual support for both formal and informal types. The formal types are as in other programming languages, which allow expressing hard constraints. The informal types (unique to MindScript) on the other hand offer flexible inductive constraints, similar to how our observations guide our own thought processes.

In practice, formal computation within MindScript is realized through a Turing-complete language (λ), while informal computations are handled by an oracle, realized through a language model (Ψ) which interprets and processes less structured inputs (hence λ^Ψ).

Features

• Implements an oracle machine.

• Minimal C-like/JavaScript/Lua syntax on top of JSON data types.

• Everything is an expression.

• The formal type system is:

  • dynamic (runtime-checked),
  • structural (based on the properties, not on names or object hierarchies),
  • and strong (type rules are strictly enforced).

• Code comments are informal type annotations.

• (Current version) interpreter implemented in Python.

Applications

• Applications that use large language models
• Processing of unstructured information
• Language model agents
• Semantic web
• and much more.

Requirements

• Python 3.9 or higher
• an LLM backend (see installation)

Disclaimer

This is a strictly experimental programming language. The Python implementation does not aim to be efficient and most likely contains bugs.