From code to images, from languages to documents.

Technology connecting code analysis, images, typesetting and languages.

Understand, improve, and transform code on a deterministic analysis foundation.

Code analysis

Lena Code

Code analysis, built on a deterministic pipeline.

Goals

  • A workspace to understand, improve, and transform code across 30 languages
  • Connecting Lena and Facet IR analysis, existing engine conversions and images
  • Project-level analysis and migration
Input code passes through a deterministic analysis foundation toward explanation, review and conversion, with the original and result read side by side.

Image as language

codeimage.io

All at once, rather than token by token. Generation with a global shutter.

Goals

  • A program format read directly from images
  • Generation that draws the whole at once and edits in parallel
  • A deterministic decoder that fails closed on invalid input
From generation built one line at a time to generation that draws the whole simultaneously.

Deterministic typesetting

ATLAS

The same manuscript and fonts produce the same pages.

Goals

  • From book interiors to reports, contracts, textbooks and academic documents
  • A document foundation with typesetting rules written in a language
The rules behind a book spread extend into grids for reports, contracts, textbooks and academic documents.

Languages

Topaz · Lispex

We refine our languages through our products.

Goals

  • Dogfooding through ATLAS, Jeniapp, Kakao Games, codeimage.io and sicp.io
  • Execution performance comparable to Rust
  • Publishing the language for publishing decisions
Languages shape products. Experience from those products refines the languages in turn.

Computer science education

sicp.io · compiler.co.kr

Learn by running code and following where it goes.

Goals

  • SICP in five chapters and 105 lessons, run in a Lispex workbench
  • The Life of Code in six chapters, from tokens to self-hosting
  • A code village of First Steps, the Lab, the Language Tour and the Field Guide
  • Examples shown with results the Topaz engine actually produced
The code 1 + 2 splits into tokens, joins into a syntax tree and becomes the value 3.

Core Semantic Kernel

Principles in papers, reproduction in public, application in products.

Papers, public reproduction and product application form the research foundation beneath the five strands.