Skip to content

Terse

Law I

"Knowledge is not stored, it is organized. Retrieval is not lookup, it is reconstruction."

Law II

"Capability is not authorization."

Law III

"The compiler works harder so you don't have to."

Terse is an original programming language designed for AI systems — conceived and built by Lesley Ancion of Goatface Tech. It reads like English, compiles to native machine code via LLVM, and treats compression, inference, knowledge graphs, tensors, and ethics as first-class language primitives rather than libraries.

Every language that exists was designed for general computing, then adapted for AI. Terse is different. It was designed for AI from day one.

What Makes Terse Different

Feature Most Languages Terse
Inference Library call Language keyword
Compression Manual / library First-class type
Knowledge graphs Build it yourself Native primitive
Sequence learning External ML library Built-in via learn / predict
Tensors Import NumPy Native type
Ethics enforcement Bolted-on filter Language construct → hardware chip
AI primitives Bolted on Core grammar
Speed Interpreted or GC Native via LLVM
Type system Often dynamic Statically typed
Readability Syntax-heavy Reads like English

Syntax Preview

// Knowledge and inference
know dog is animal
know dog has fur
when has fur then is mammal
infer dog

// Variables, math, conditionals
x = 10
y = x * 2
if y > 15
  know result is large

// Strings
greeting = "hello world"
size = length(greeting)
shouted = replace(greeting, "world", "WORLD")

// Tensors
weights = tensor 3 3 fill 1.0
output = weights dot inputs

// Ethics — cryptographically sealed
sealed ethics_core
  ethics rule no_harm
    when intent is harm
    then deny with reason: "Law II violation"

// Loops and functions
to classify thing
  infer thing
  return thing

each creature in animal
  classify creature

Project Status

  • ✅ Session 1 — Lexer & Parser
  • ✅ Session 2 — Interpreter, Knowledge Base & Markov Chains
  • ✅ Session 3 — Loops, main.py & First Running Program
  • ✅ Session 4 — REPL & Interactive Prompt
  • ✅ Session 5 — Variables, Math & Conditionals
  • ✅ Session 6 — Tensors, Compression & Ethics Rules
  • ✅ Session 7 — Semantic Compression Algorithm & Sealed Blocks
  • ✅ Session 8 — C Transpiler & IR Pipeline
  • ✅ Session 9 — First Native Binary & NCI Integration
  • ✅ Session 10 — Phase 3 Complete — LLVM Compiler, Inference, and Ethics Running Natively
  • ✅ Session 11 — Phase 3.1 Complete — Float Support
  • ✅ Session 12 — Phase 3.2 Complete — String Manipulation & Static Typing
  • ✅ Session 13 — Phase 3.2.5 Complete — Lists, Escape Sequences, and Compiler Improvements

Phase 1 — Python Interpreter: complete ✅
Phase 1.5 — Semantic Compression: complete ✅
Phase 1.6 — Runtime Integrity & Encryption: complete ✅
Phase 2 — C Transpiler: complete ✅
Phase 2.5 — First Native Binary: complete ✅
Phase 3 — LLVM Compiler: complete ✅
Phase 3.1 — Float Support: complete ✅
Phase 3.2 — String Manipulation: complete ✅
Phase 3.2.5 — Lists: complete ✅

The Vertical Stack

Three original works by Lesley Ancion under the Goatface Tech brand:

  • NCI — Native Compression Intelligence. The AI architecture Terse is the native language of.
  • Terse — The language. Every AI primitive is core grammar, not a library.
  • NCI Ethics Core — A dedicated hardware chip that executes Terse ethics rules in silicon. You can't jailbreak hardware.

The language, the AI, and the silicon — designed together from day one.

Author

Lesley Ancion · Goatface Tech · Sundre, Alberta · 2026

Concept disclosed and timestamped. See concept disclosure PDF.