Smallsome

Welcome

Welcome to this research project. Smallsome brings together encoders and decoders that I developed in 2026 with the help of ChatGPT. Through many iterations, my goal was to create formats that are very small, very fast and still powerful. Encoding was intended to take longer than decoding, and we settled on four broad quality levels: lossless, medium, small and junk.

My Unicode-based programming language, Cymple, is now called Smallsome Symbols. Version 3.0 introduces an ASCII mode, Smallsome Code. Finally, the project includes a theoretical µRISC CPU tailored to these formats, together with a matching theoretical computer specification.

Without ChatGPT, I could not have achieved this.

Current Versions: RC 26, Smallsome Code 3.0, µRISC 1 (rev 4)

Jörg Burbach, October 10, 2026 - joerg-burbach.de

One language, two source modes

fn sum_to(n: num) -> num
    num total = 0
    for i in 1..n
        total = total + i
    return total

num answer = sum_to(8)
print "Sum: {answer}"

🧵 sum_to(n: 🔢) -> 🔢
    🔢 total ← 0
    🔁 i in 1..n
        total ← total + i
    ↩ total

🔢 answer ← sum_to(8)
💬 "Sum: {answer}"

Same types, ownership, control flow and runtime contract. A source file selects one mode. num retains binary64 semantics; word explicitly requests 32-bit wrapping arithmetic.

Language specificationShared grammar

Reference formats

RAU, RFXL, RFXA, PMF0, RPMC, LeXA, FormA and WORM0. Defined byte streams, memory-only APIs and existing source references.

CPU integration is theoretical. A codec reference is not evidence of measured T1 decoding performance.

All formatsRFXL reference image

The theoretical machine

Project status