Homogeneous data
Typed vectors, nested vectors, matrices, and numerical kernels.
[1 2 3 4]Neo 3.1.0Native compiler and Neo interpreter
A vector-native language.
Neo applies the same operations to scalars, vectors, and matrices. It has first-class unknown values and a complete interpreter written in Neo.
3.1.0 current releaseNeo in Neo interpreter milestone
example.neoNeo 3.1
{def square (x) {* x x}} {square 5} {square [1 2 3 4]}
output
25
[1 4 9 16]Current compiler syntax and behavior.
Philosophy
Programs transform vectors.
Vectors are the basic unit of data and computation in Neo.Interpreter written in Neo
Neo can now run through Neo.
The repository includes a complete interpreter implemented in the language itself. It has its own lexer, parser, evaluator, runtime environment, file runner, and persistent multiline REPL.
It executes Neo's full regression suite. The native compiler remains the path to fast executables; the interpreter is the semantic reference and the next step toward self-hosting.
Neo REPLimplemented in Neo
Neo REPL. Type :quit to exit.
> {var x 40}
40
> {+ x 2}
42
> {def square (value)
... {* value value}}
> {square 6}
36NEO FULL REGRESSION TEST PASSED
Syntax
The brackets tell you what you are looking at.
Square brackets hold typed data, parentheses hold mixed data, and braces run code.
Heterogeneous data
Dictionaries, mixed values, complex numbers, and stored programs.
(name:"Neo" version:3)Executable code
Calls, functions, declarations, control flow, and scoped blocks.
{+ 1 2}Vector operations
Write the function once. Use it on a number or a vector.
Scalars broadcast. Vectors are handled element by element. Ordinary arithmetic needs no separate array API.
No
mapNo foreachNo manual indexing{def affine (x scale bias) {+ {* x scale} bias}} {affine 5 2 1} ; 11 {affine [1 2 3 4] 2 1} ; [3 5 7 9]
scalar
{square 5}25vector
{square [1 2 3 4]}[1 4 9 16]broadcast
{+ [1 2 3] 10}[11 12 13]Unknown values
Missing data has a value: ?
? can live inside ordinary vectors and participate in arithmetic, comparisons, logic, statistics, and control flow. Errors use the separate value !.
?
unknown or missing value
!
runtime error value
if
true, false, and unknown branches
{var temperature ?} {print {if {< temperature 40} "safe" "too hot" "sensor unavailable"}}
output
sensor unavailableSpreadUse
Use @ when you need every combination.
Normal vector calls are element-wise. Mark an input with @to request a Cartesian product. Two marked vectors produce a matrix; more naturally produce higher-dimensional results.
102030
12
112131122232
{+ @[1 2] @[10 20 30]}Standard library
Numerical tools are available from the start.
Statistics, matrices, AI primitives, complex numbers, quaternions, dictionaries, JSON, file I/O, and set operations share Neo's vector model.
Data analysis
filter · reduce · sort · find · JSON · dictionariesStatistics
mean · variance · sd · min · maxMatrices
matmul · transpose · outer · identity · traceAI primitives
relu · sigmoid · softmax · mse · cross-entropyNumerical types
complex · quaternion · conjugate · rotationPrograms and text
command-line args · string concat · string push/pop · file I/O{def layer (input weights bias) {relu {+ {matmul input weights} bias}}}
“Analyze the data, then compile the program.”
Read data, transform vectors, handle missing values, compute statistics, perform matrix operations, and ship the result as a native application.
Made for numerical and systems work.
Robotics
Sensor processing, control systems, matrices, quaternions, and hardware libraries.
Edge AI
Compact inference pipelines, activations, loss functions, and native deployment.
Games
Vector math, simulation, rotation, procedural systems, graphics and engine libraries.
Data
Statistics, JSON, dictionaries, unknown values, and executable transformation pipelines.
Embedded
Device interaction, signal processing, small processors, and deterministic memory work.
Code and data
Store code as data, then run it when you choose.
Heterogeneous data vectors keep code inert. The {run}bridge makes the move from data to execution visible, allowing programs to be generated, extended, and executed using the same vector principles.
Neo 3.1.0