MOL¶

The pipeline language for AI, data, and automation — with auto-tracing built in.

The Problem¶

Every pipeline — AI, data, or automation — is invisible glue code. You chain operations together, then add print() and logging everywhere to see what happened.

MOL fixes this at the language level.

The Solution¶

let index be doc |> chunk(512) |> embed("model-v1") |> store("kb")
let answer be retrieve(query, "kb", 3) |> think("answer this")
guard answer.confidence > 0.5 : "Low confidence"

Run it, and MOL automatically shows you:

  ┌─ Pipeline Trace ──────────────────────────────────────
  │ 0.  input                   ─  <Document "data.txt" 339B>
  │ 1.  chunk(512)          0.1ms  → List<5 Chunks>
  │ 2.  embed("model-v1")   0.2ms  → List<5 Embeddings>
  │ 3.  store("kb")         0.0ms  → <VectorStore "kb" 5 vectors>
  └─ 3 steps · 0.4ms total ───────────────────────────

No print statements. No logging. Auto-tracing is built into |>.

:material-pipe: Pipe Operator `|>`¶

Chain operations left-to-right with automatic execution tracing. Every stage is timed and typed.

"hello" |> upper |> split(" ")

:material-shield-check: Guard Assertions¶

Runtime validation built into the language. No try/catch boilerplate.

guard confidence > 0.5 : "Too low"

:material-brain: Domain Types¶

12 built-in types: Thought, Memory, Node, Document, Chunk, Embedding, VectorStore, and more.

:material-function: 210 Stdlib Functions¶

Math, strings, lists, maps, sorting, searching, JSON, vectors, encryption, JIT tracing, swarm computing, and RAG operations.

:material-translate: Transpiler¶

Write once in MOL, transpile to Python or JavaScript.

mol transpile app.mol --target python

:material-security: Access Control¶

Built-in security model for AI resource management.

access "mind_core" with "admin"

🚀 v2.0 — Kernel-Grade Evolution¶

Five new systems transform MOL into infrastructure for Neural Kernel and De-RAG / Sovereign Memory.

:material-shield-lock: Memory Safety¶

Rust-inspired borrow checker: own, borrow, transfer, release, lifetime. Zero buffer overflows, zero use-after-free — enforced at the language level.

let own buffer be [1, 2, 3]
let ref view be borrow buffer
lifetime scope do ... end

:material-vector-line: Native Vectors¶

First-class Vector type with SIMD-like ops, ANN search, quantization. 25 functions for nanosecond retrieval.

let v be vec(1.0, 0.0, 0.0)
vec_cosine(v, vec_rand(3))
vec_index_search(idx, query, 5)

:material-lock: Integrated Encryption¶

Homomorphic encryption (Paillier), symmetric crypto, zero-knowledge proofs. Compute on ciphertext without decrypting.

let keys be crypto_keygen(512)
let enc be he_encrypt(42, keys)
he_add(enc, he_encrypt(18, keys))

:material-lightning-bolt: JIT Tracing¶

Self-optimizing hot-path detection. Type specialization, inline caching, automatic recompilation.

show jit_stats()
show jit_hot_paths()

:material-lan: Swarm Runtime¶

Multi-node distributed execution. Consistent hashing, data sharding, MapReduce, dynamic scaling.

let c be swarm_init(5, 2)
swarm_shard(data, c, "hash")
swarm_map(c, fn(d) -> len(d))

Built For Real Domains¶

MOL is purpose-built for domains where pipeline visibility and readable code directly reduce debugging time.

:material-brain: AI & ML Pipelines¶

Build RAG pipelines in 3 lines. Every stage auto-traced with timing and type info. Built-in domain types: Document, Chunk, Embedding, VectorStore.

doc |> chunk(512) |> embed |> store("kb")

:material-chart-bar: Data Processing & ETL¶

Smart functions eliminate boilerplate. filter("active"), sort_by("name"), pluck("id") — no lambdas needed for common operations.

sales |> filter("closed") |> sort_by("amount") |> pluck("rep")

:material-server: DevOps & Automation¶

Log analysis, monitoring, SLA validation with guard assertions and built-in statistics (mean, median, percentile).

logs |> pluck("latency") |> percentile(95)

See All Use Cases

What Makes MOL Different¶

Feature	Python	Elixir	F#	MOL
Pipe operator `\\|>`	—	✓	✓	✓
Auto-tracing	—	—	—	✓
AI domain types	—	—	—	✓
RAG built-in	—	—	—	✓
Guard assertions	—	✓	—	✓
Transpiles to Python & JS	—	—	—	✓
Built-in algorithms	—	—	—	✓

Built By¶

CruxLabx — Building the cognitive infrastructure for IntraMind.

MOL: Where pipelines think.