Skip to content
pnxt

Future Vision

The pnxt paradigm has realized many of its original theoretical goals through Phases 4–7. The roadmap now extends toward self-modification, self-hosting, and ecosystem growth.

Completed Goals

The following goals from the original vision have been fully implemented and verified:

  • Bridge Grammar — Constrained-decoding JSON Schema forcing LLMs to output valid VPIR nodes, with auto-repair and confidence scoring (Sprints 2, 11, 12)
  • VPIR Compiler/Interpreter — Full execution engine with parallel wave scheduling, result caching, and DPN runtime compilation (Sprints 3–4, 10)
  • Natural Language Protocol Design — Three protocol state machines (delegation, negotiation, resolution) over DPN channels (Sprint 3)
  • Tree-sitter DKB Integration — Knowledge graph with 8 entity kinds, 8 relation types, traversal, and HoTT conversion (Sprint 5)
  • LLMbda Calculus — Typed lambda calculus with IFC, noninterference, and VPIR semantic bridge (Sprint 3 of Phase 6)
  • SMT Solver Integration — 21 formally verified Z3 properties plus CVC5 multi-solver support (accumulated across Sprints 2–12)
  • Dataflow Process Network Engine — Typed FIFO channels, process actors, DPN runtime, supervisor pattern, bisimulation (Sprints 1, 4, 7, 10)
  • HoTT Typed Tokenization — Categories, morphisms, n-paths, univalence, transport, categorical tokenizer experiment (Sprints 5–6, 9)
  • Neurosymbolic Bridge — P-ASP confidence scoring, Active Inference graph patching, refinement pipeline (Sprint 8)
  • Self-Description (M1) — pnxt describes, validates, categorizes, and executes its own pipeline as VPIR (Sprint 9)
  • External Task Expression (M2) — Real-world tasks expressed entirely in VPIR JSON, no TypeScript required (Sprint 11)
  • LLM-Native Programming Foundation (M3) — Reliable generation pipeline with error taxonomy, auto-repair, and confidence scoring (Sprint 12)

Near-Term Goals (Phase 7 Remaining)

M4: Self-Modification

pnxt modifies its own pipeline through VPIR:

  • Pipeline stages themselves become VPIR-editable
  • Agents propose optimizations as VPIR graph transformations
  • Changes verified via Z3 before application

M5: Self-Hosting

pnxt’s core components expressed in pnxt itself:

  • Transition from TypeScript host to pnxt-native execution
  • Categorical syntax replaces JSON intermediate format
  • Bootstrap compiler validates the paradigm end-to-end

Long-Term Goals (Phase 8+)

Web-Based Visualization Frontend

Interactive node-graph renderer consuming the JSON export format:

  • VPIR graph visualization with execution trace overlay
  • HoTT category browser with morphism navigation
  • Pipeline stage flow with real-time execution monitoring

Multi-Language Tree-sitter Parsers

Extend KG parsing beyond TypeScript:

  • Python, Rust, Go, and Java parsers
  • Cross-language dependency analysis
  • Unified knowledge graph across polyglot codebases

Categorical Token Embeddings

Transformer fine-tuning with morphism-structured embeddings:

  • Build on the categorical tokenization experiment (42-token vocabulary)
  • Train embeddings that preserve categorical structure
  • Evaluate LLM performance on morphism-aware tokenization

Distributed DPN

Multi-node actor execution for scale:

  • Network-transparent channel communication
  • Distributed supervisor trees
  • Cross-node IFC enforcement

Community & Ecosystem

Open standards and adoption tooling:

  • Open specification for ANP protocols
  • Reference implementations in multiple languages
  • Developer tooling and IDE integrations
  • Community governance and contribution model