Agent-Computer Interface
Structured protocols for agent-to-system communication with versioned capabilities, clear error handling, and recovery patterns. Agents negotiate what they can do, not just what they’re told.
pnxt — Agent-Native Programming
A net-new programming paradigm built exclusively for LLMs and AI agents.
Moving beyond human-readable syntax toward structured, verifiable, graph-based program representations.
Why a New Paradigm?
Traditional programming languages are optimized for human cognition — visual hierarchy, short-term memory, lexical parsing. LLMs excel at structural data manipulation (JSON/graphs) but struggle with implicit control flow and loop-state tracking.
pnxt designs an execution environment where LLMs orchestrate logic graphs rather than generate syntax. In the Agent-Native Programming (ANP) paradigm, AI agents are first-class entities with identity, memory, state, and tools — not just coding assistants bolted onto existing workflows.
The Three Pillars
Semantic Memory
Persistent, evolving memory — working, semantic, and episodic — that makes agents more valuable in their hundredth session than their first. Memory is the foundation of coherent agency.
Natural Language
Natural language as the primary interface for task specification, progress reporting, and collaboration. Human partnership, not replacement.
Theoretical Foundations
HoTT Typed Tokenization
Code as categorical objects, morphisms, and paths — built on Homotopy Type Theory for mathematical rigor.
VPIR
Verifiable Programmatic Intermediate Representation with mechanically verifiable reasoning chains.
Dataflow Process Networks
Actors communicating via FIFO channels, eliminating imperative loops in favor of declarative data flow.
LLMbda Calculus
Lambda calculus with Information Flow Control for noninterference guarantees.
SMT Solvers
Z3/CVC5 for constraint satisfaction and formal verification of agent behaviors.
Bridge Grammar
Constrained-decoding JSON schema forcing LLMs to output valid VPIR nodes.
Key Principles
Incrementalism Every component is designed for phased adoption — start small, grow organically.
Structural Safety Correct behavior is easy by design. Incorrect behavior is structurally difficult.
Explicit Over Implicit Side effects declared, capabilities negotiated, trust measured — nothing hidden.
Memory as Foundation Persistent memory transforms stateless interactions into coherent agent experiences.
Project Status
| Phase | Focus | Status |
|---|---|---|
| Phase 1 | Core Architecture, State Separation & FFI | Complete |
| Phase 2 | Bridge Layer & Mathematical Spec | Complete |
| Phase 3 | Deep analysis of pillars, patterns, and architecture | Complete |
| Phase 4 | Infrastructure prototype & empirical evaluation | Complete |
| Phase 5 | Paradigm Foundation (DPN, VPIR, Bridge Grammar, HoTT, Z3) | Complete |
| Phase 6 | Integration & Deepening (9 sprints — self-hosting, categorical tokenization) | Complete |
| Phase 7 | Self-Hosting Paradigm (handler library, VPIR authoring, reliable bridge grammar) | In Progress |
View Full Project Status & Roadmap See the current state of the project, completed milestones, and what's coming next.