Skip to content

Changelog

All notable changes to MOL are documented here.

[2.0.1] - 2026-02-23

Security

  • Fixed Full RCE vulnerability — Python class hierarchy traversal via dunder attributes (__class__.__subclasses__()) allowed arbitrary command execution on the playground server. Blocked all __-prefixed attribute and method access in the interpreter. Applied in all modes (sandbox + local) for defense in depth.
  • Added 11 security regression tests (213 total tests pass)

Security Credits

  • a11ce — Reported the Full RCE vulnerability via responsible disclosure. Thank you for helping keep MOL safe.

[2.0.0] - 2026-02-21

Added — Kernel-Grade Evolution (56 new functions, 5 new systems)

🛡️ Memory Safety — Borrow Checker (borrow_checker.py)

  • Rust-inspired ownership model: own, borrow, borrow_mut, transfer, release
  • Lifetime scopes with automatic resource cleanup
  • Use-after-free detection, buffer overflow prevention
  • AI-assisted safety analysis
  • New AST nodes: OwnDeclare, BorrowDeclare, BorrowMutDeclare, MoveOwnership, DropValue, LifetimeScope

📐 Native Vector Engine (vector_engine.py) — 25 functions

  • First-class Vector type backed by array.array for SIMD-like performance
  • Arithmetic: vec_add, vec_sub, vec_scale, vec_concat
  • Similarity: vec_dot, vec_cosine, vec_distance, vec_normalize
  • Factory: vec, vec_zeros, vec_ones, vec_rand, vec_from_text
  • ML: vec_softmax, vec_relu, vec_quantize (int8 compression)
  • ANN search: vec_index, vec_index_add, vec_index_search with LSH bucketing
  • Batch operations: vec_batch_cosine, vec_top_k

🔐 Integrated Encryption (encryption.py) — 15 functions

  • Homomorphic encryption (Paillier scheme): crypto_keygen, he_encrypt, he_decrypt, he_add, he_sub, he_mul_scalar
  • Symmetric encryption (HMAC-based): sym_keygen, sym_encrypt, sym_decrypt
  • Zero-knowledge proofs: zk_commit, zk_verify, zk_prove
  • Utilities: secure_hash, secure_random, constant_time_compare

⚡ Self-Optimizing JIT Tracing (jit_tracer.py) — 7 functions

  • Hot-path detection with configurable thresholds (50/100 calls)
  • Type profiling and specialization (Int, Float, String fast-paths)
  • Inline caching and constant folding
  • Functions: jit_stats, jit_hot_paths, jit_profile, jit_reset, jit_warmup, jit_enabled, jit_toggle

🌐 Multi-Node Swarm Runtime (swarm_runtime.py) — 12 functions

  • Consistent hash ring for data distribution
  • Functions: swarm_init, swarm_shard, swarm_map, swarm_reduce, swarm_gather
  • Communication: swarm_broadcast, swarm_scatter
  • Management: swarm_health, swarm_nodes, swarm_rebalance, swarm_add_node, swarm_remove_node
  • ThreadPoolExecutor-based parallel execution with fault tolerance

Infrastructure:

  • Grammar updated with ownership syntax rules
  • Parser extended with 6 new transformer methods
  • Interpreter extended with 7 new exec methods + JIT integration in all function calls
  • Type system extended: Vector, QuantizedVector, VectorIndex, EncryptedValue, EncryptedVector, EncryptedMemory, CryptoKeyPair, SwarmCluster
  • Total stdlib: 210 functions (up from 162)
  • 6 new examples: 20_vector_engine, 21_encryption, 22_memory_safety, 23_jit_tracing, 24_swarm_runtime, 25_sovereign_pipeline
  • All 168 existing tests pass with zero regressions

[0.3.0] - 2026-02-10

Added

Universal Algorithms & Functions (42 new functions):

  • Functional programming: map, filter, reduce, find, every, some, group_by
  • List operations: flatten, unique, zip, enumerate, count, find_index, take, drop, chunk_list
  • Math: floor, ceil, log, sin, cos, tan, pow, clamp, lerp, pi, e
  • Statistics: mean, median, stdev, variance, percentile
  • String algorithms: starts_with, ends_with, pad_left, pad_right, repeat, char_at, index_of, format
  • Hashing & encoding: hash (SHA-256/MD5/SHA-1/SHA-512), uuid, base64_encode, base64_decode
  • Sorting: sort_by, sort_desc, binary_search
  • Random: random, random_int, shuffle, sample, choice
  • Map utilities: merge, pick, omit
  • Type checking: is_null, is_number, is_text, is_list, is_map
  • Output: print (multi-argument)

Infrastructure:

  • User-defined functions (MOLFunction) are now first-class Python callables — can be passed to map, filter, reduce, etc.
  • MkDocs Material documentation site with 20+ pages
  • Total stdlib: 90+ functions
  • Tests: 68 passing (25 new)

[0.2.0] - 2026-02-09

Added

  • Pipe operator |> — chain operations left-to-right
  • Auto-tracing — pipe chains of 3+ stages automatically print timing and type info
  • Guard assertions — guard condition : "message"
  • Named pipelines — pipeline name(params) ... end
  • RAG types: Document, Chunk, Embedding, VectorStore
  • 15 new functions: load_text, chunk, embed, store, retrieve, cosine_sim, think, recall, classify, summarize, display, tap, assert_min, assert_not_null, Chunk constructor
  • CLI flags: --trace / --no-trace
  • New examples: 07_pipeline.mol, 08_rag_pipeline.mol
  • VS Code extension updated with new keywords and functions
  • Tests: 43 passing (22 new)

[0.1.0] - 2026-02-08

Added

  • LALR parser with Lark grammar (131 rules)
  • Core interpreter with visitor pattern
  • Types: Number, Text, Bool, List, Map, null
  • Domain types: Thought, Memory, Node, Stream
  • Control flow: if/elif/else, while, for
  • Functions with closures and recursion
  • Event system: trigger, listen, emit
  • Neural primitives: link, evolve, process
  • Access control security model
  • Type annotations: let x: Type be value
  • Python transpiler
  • JavaScript transpiler
  • CLI: run, parse, transpile, repl, version
  • VS Code extension with TextMate grammar + snippets
  • 21 tests, 6 examples, 6 tutorials
  • Comprehensive documentation