GRISP is a graph-rewrite symbolic programming language designed from the ground up for the age of AI-assisted software development.
It is:
- AI-native — G-Blocks syntax is easy for LLMs to generate and reason about
- Deterministic — execution is pure graph rewriting
- Explainable — rules are transparent, inspectable, and composable
- Safe — no arbitrary code execution, only structural transformations
- Extensible — new rules = new capabilities
GRISP 1.0 includes:
- A full G-Blocks parser
- A complete graph engine (
TGNode,TGEdge,TGGraph,TGValue) - A pattern matcher with backtracking and variable binding
- A rewrite engine with node/value substitution and splicing
- A deterministic runtime loop
- A full test suite
GRISP is implemented in Delphi for clarity, speed, and strong typing.
Modern AI systems generate text — but text is not execution.
GRISP provides the missing layer:
A deterministic, symbolic execution substrate for AI-generated logic.
GRISP is ideal for:
- AI-generated compilers
- AI-generated optimizers
- DSLs and meta-languages
- Knowledge graph reasoning
- Program synthesis
- Automated theorem proving
- Static analysis
- AI-assisted refactoring
- Symbolic computation
node rule.swap_if_greater {
match: node = {
X: node = {
value: number = 2
next: identifier = Y
}
Y: node = {
value: number = 1
}
}
rewrite: node = {
X: node = {
value: number = 1
next: identifier = Y
}
Y: node = {
value: number = 2
}
}
}