A premier systems and AI programming language created by Rohit Pathak at TEZZCORP PRIVATE LIMITED. Compiling directly to zero-overhead standalone native executables with first-class tensor arithmetic, multi-threaded coroutines, and scoped memory safety without garbage collection pauses.
irm https://tezznative.org/install.ps1 | iex
tezz init my-ai-project && cd my-ai-project
tezz add mind && tezz sync
tezz build --release
TezzNative compiles directly into native CPU machine instructions (AVX2, FMA, NEON) with zero garbage-collection pauses and zero interpreter overhead.
Explore interactive code examples showcasing concurrent actor tasks, dense matrix math & softmax, INT8 quantized inference, systems bump allocator, and high-throughput HTTP routing.
// Concurrent Worker Tasks & Mailboxes in TezzNative import "actor" import "net" fn handle_worker(sys:*ActorSystem, worker_id:int, master_id:int, uid:int) -> int: say "Spawning worker task for UID: ", uid resp:str = "HTTP/1.1 200 OK\r\nContent-Type: application/json\r\n\r\n{\"user\":\"ok\"}" status:int = net.http_status_code(resp) // Send status back to master actor mailbox ret actor.send_from(sys, worker_id, master_id, 1, uid, status, 0, 0, "ok") fn main() -> int: say "Spawning concurrent actor worker tasks..." sys:*ActorSystem = actor.system_new(4, 4, 1) if sys == 0: ret 1 master:int = actor.spawn(sys, "master") w1:int = actor.spawn(sys, "worker-1") w2:int = actor.spawn(sys, "worker-2") handle_worker(sys, w1, master, 101) handle_worker(sys, w2, master, 102) msg1:ActorMessage msg2:ActorMessage unsafe: actor.receive(sys, master, &msg1) actor.receive(sys, master, &msg2) say "Worker 1 HTTP status: ", msg1.arg2 say "Worker 2 HTTP status: ", msg2.arg2 actor.system_free(sys) ret 0
TezzNative eliminates the historic divide between rapid Python prototyping and low-level C systems programming.
Write elegant indentation-based code that feels effortless to read, while enjoying compile-time type safety, exhaustive pattern matching, and zero runtime type tags.
fn fib(n: int) -> int:
if n <= 1: ret n
ret fib(n - 1) + fib(n - 2)
Spawn hardware-threaded async workers and await asynchronous I/O and matrix compute non-blockingly on an integrated event loop with zero function-color baggage.
let task: AsyncTask = task.spawn_arg(worker, 42)
let result: int = await task
Tensors are first-class language constructs. Execute matrix multiplications with the infix @ operator, slice 4D tensor volumes seamlessly, and backpropagate with automatic differentiation.
let H = (X @ W1).relu()
let slice = tensor[0, 1, 10:20, 10:20]
Resource handles, sockets, files, and CUDA contexts are automatically released in LIFO reverse order upon scope exit. Zero garbage collection pauses, zero memory leaks.
let f = io.open("data.bin", "rb")
defer f.close() // Executed on exit
Compile single standalone Windows PE .exe or Linux ELF binaries that require zero runtime installs, zero dynamic interpreters, and start in under 1 millisecond.
tezzc buildexe main.tn app.exe
# Outputs standalone native binary
Official LSP v3.17 daemon (tezz_lsp.exe) brings hover documentation, Go to Definition, auto-completion, and real-time syntax checking to VS Code, Neovim, Zed, and Cursor.
tezzc lsp # Starts standard stdio JSON-RPC daemon
TezzNative includes a first-party Language Server Protocol daemon (tezz_lsp) built directly in TezzNative. Enjoy rich IDE features with zero configuration across your favorite code editor.
import "tztensor"
import "tzautodiff"
fn forward_pass(x: Tensor, w: Tensor):
let y_hat = (x @ w).relu()
ret y_hat
Join developers creating high-throughput AI backends, native desktop GUIs, and zero-overhead systems software.