Getting Started
Vox: The AI-Native Programming Language
Getting Started with Vox
Journeys
Journey: Building Resilient AI Agents
Journey: Reliable Background Workflows
Journey: One-File Full-Stack Data
Journey: Native Rust LLM Training
Tutorials
Tutorial: Building UI with Islands
Tutorial: Building a Collaborative Task List
Tutorial: Persistent Actors & State
Tutorial: Workflow Durability
Tutorial: first .vox app (checkpoints)
How-To Guides
@py.import – Python Library Integration (torch, numpy, etc.)
Contributing — Mens native training
Contributing — Populi control plane
Contributing — parser and HIR
Ecosystem & Tooling
Examples
First Full Stack App
Golden Examples Corpus
How To: Train Mens on RTX 4080 Super
How to use the canonical VoxDB / Codex store
How-To: Build AI Agents and MCP Tools
How-To: Deploy to Production
How-To: Handle Errors Gracefully
How-To: Islands and Pages
How-To: Model Complex Domain Logic
How-To: Publish Scientia findings
How-To: Rust crate imports in Vox scripts
How-To: Scale Actors
How-To: System I/O and Capabilities
How-To: Test Your Logic
How-To: Testing Integration
How-To: The Database Layer
Model Routing & Provider Cascade
Scientia publication: what you type vs what the system derives
Troubleshooting FAQ
API Reference — Crates
Known Documentation Gaps & Backlog
Crate API: vox-ast
Crate API: vox-codegen-rust
Crate API: vox-codegen-ts
Crate API: vox-dei-sandbox
Crate API: vox-gamify
Crate API: vox-hir
Crate API: vox-lexer
Crate API: vox-mcp
Crate API: vox-orchestrator
Crate API: vox-parser
Crate API: vox-py
Crate API: vox-typeck
Crate API: vox-wasm
Examples
Golden Examples: Working Vox Code
Explanations
AI Agent Orchestration
Actors & Workflows
Compiler Architecture
Explanation: Capabilities
Explanation: Compiler Lowering Phases
Explanation: Durable Execution
Explanation: Security Model
Explanation: The Vox Runtime
Glossary: Vox Terminology
Native ML Training Pipeline
OpenClaw Competitive Analysis
Rosetta Inventory: One Scenario, Four Languages
Vox FAQ: Frequently Asked Questions
Why Vox: Compiler-Verified AI Code
README: Vox Platform (Scientia Draft, April 2026)
Architecture Decisions (ADRs)
ADR 002 — Diátaxis Three-Tier Documentation Architecture
Architecture SSOTs
Architecture index
Compiler diagnostics and Rust codegen ergonomics
Cross-repo querying and observability
Language surface SSOT (keywords, decorators, manifests)
OpenAPI contract SSOT (Populi, MCP, Codex)
Outbound HTTP policy (reqwest / vox-reqwest-defaults)
Vox source → compiler → Mens training (pipeline SSOT)
Populi data pipeline (control plane vs Mens corpus)
AI CLI Generation Standard
Capability registry SSOT
Capability visualization views
MCP exposure from the Vox language (SSOT)
Additive schema plan: scholarly external jobs and snapshots
Anti-foot-gun planning standard
CLI design rules SSOT
CLI reachability SSOT
CodeRabbit review coverage SSOT
Compiler IR Pipeline
Completion policy SSOT (LLM premature-completion)
Dependency Sprawl Audit and Resolution (2026)
Deployment Compose SSOT
Doc-to-code acceptance checklist
Document boundary matrix
Document maintenance protocol
Exception and deferral policy
Fast LLM instruction plan
Feature growth boundaries
God object defactor checklist (v3)
HITL Doubt Loop (SSOT)
Hybrid adapter cookbook (SPA + SSR)
IR emission SSOT (HIR, WebIR, VoxIrModule)
Internal Web IR Implementation Blueprint
Internal Web IR Side-by-Side Schema
Interop tier policy
Legacy retirement roadmap (2026)
Ludus / gamify schema inventory (SSOT pointers)
Ludus: scope and non-goals
Maintainability hotspot matrix (baseline)
Master planning index
Mens lane segmentation research
Mens training SSOT
Milestone and gate definition spec
Minimal React Interop Shell Strategy
Mobile/Desktop Convergence & Language Extension Research 2026
News syndication: incident patterns and mitigations
Nomenclature migration map (SSOT)
Operations catalog SSOT
Orchestrator AgentEventKind → Ludus matrix
Orchestrator multi-agent groundwork (2026)
Plan adequacy — thin plans, external limits, and Vox mitigation
Planning critique and gap analysis
Planning meta exception register
Planning meta maintenance log
Planning taxonomy and glossary
Populi GPU truth probe specification (NVML Layer A)
Populi node lifecycle, drain, and GPU hotplug
Question gate standard for planning
Qwen 3.6 integration research (groundwork)
Qwen3.5 Multimodal Phase 2 Backlog
React interop full-repo migration charter (2026)
React interop migration backlog (2026)
Research baseline and source-of-truth map
Rust ecosystem support SSOT
SCIENTIA A2A evidence-gathering tasks
SSOT / DRY convergence roadmap
Scaling CI enforcement rollout
Scaling audit baseline (workspace map)
Scholarly publication: digest-bound approval invariants
Script surface audit and Vox migration
TOESTUB scaling rules (SSOT)
Table metadata SSOT (Arca ↔ @table convergence)
TanStack Start Codegen Specification
TanStack Start Implementation Backlog
Task catalog authoring spec
Telemetry client disclosure SSOT
Telemetry remote sink specification
Telemetry trust boundary and SSOT map
Trust Reliability Layer (SSOT)
Unified News Syndication Security & Safety
Vox Architectural Organization & Governance
Vox Docker-backed portability implementation plan 2026
Vox Docker-backed portability research 2026
Vox Ludus integration contract (producers)
Vox Memory System
Vox RAG and Autonomous Research Architecture 2026
Vox React/v0 Interop Research Findings
Vox Security Model
Vox Session Management
Vox Web: Minimal React Interop Implementation Plan
Vox bell-curve strategy
Vox boilerplate implementation status
Vox boilerplate reduction master roadmap
Vox boilerplate research findings 2026
Vox full-stack ergonomics deep dive
Vox packaging full implementation plan 2026
Vox packaging implementation blueprint
Vox packaging research findings 2026
Vox shell operations boundaries
Vox web stack SSOT
VoxDB connection policy (SSOT)
VoxGiantia publication architecture (beginner map)
Weighted deep planning manual
Clavis V2: Full Implementation Plan (2026)
Vox Publication and Orchestration Hardening: Implementation Plan 2026
Research index
Unified Agentic Control Surface Research
Protocol convergence research 2026
VCS for agent state and artifact snapshotting research 2026
Syndication SDK Deep Research & Strangler-Fig Migration Plan 2026
SCIENTIA impact, readership, and citation-adjacent signals (research seed)
Prompt engineering, system prompts, document-skills, and SCIENTIA (research 2026)
SCIENTIA publication-worthiness and SSOT unification (research 2026)
SCIENTIA multi-platform ranking, discovery, and anti-slop SSOT (research 2026)
Mens vision and multimodal inputs (research 2026)
Mens Qwen family migration and native stack (research 2026)
TOESTUB line limit and MENS corpus size research (2026)
Vox corpus lab: mass examples, metrics, and eval harness (research 2026)
2026 State-of-the-Art: Dynamic Agentic Planning & Orchestration
AI Agent Context and Handoff Research
AI IDE feature research findings 2026
AI-Augmented Testing & Hourglass Architecture Research (2026)
Agent Mesh Economics & Token Costs
Agent Trust Reliability Evaluation
Architecture Decision Checklist for Implementing Agent Handoff Continuity
Architecture: ASR Speech-to-Code
Automated Testing Research for the Vox Language
Catastrophic Forgetting in QLoRA Fine-Tuning
Clavis as a one-stop secrets manager: research findings 2026
Clavis secrets, env vars, and API key strategy research 2026
Cognitive Science and NLP: Constraint as Guide vs. Output Space Collapse
Compiler Testing Research Synthesis
Context management research findings 2026
Continual Learning Flywheel Risks
Cross-Agent Evidence Sharing in A2A Protocol Implementations
Design Pattern Recommendations for Platform Gaps
Diagnostic questioning — research synthesis 2026
Documented Failure Modes: Context Bleed and Session Identity Confusion
Empirical Evidence for Context Compaction Strategies
Empirical Evidence: Strictly-Typed vs. Dynamically-Typed Languages
Empirical Justification for Reward Weight Allocations in Code RL
Evaluating AI Plan Adequacy Heuristics
Evidence Base for Context Retrieval Policies
Execution Time Budgeting and Agent Learning Research 2026
GRPO Reward Shaping for Code LLMs
GRPO and VRAM Efficiency: Architectural Comparisons and Small-Batch Dynamics
Gap Analysis and Recommended Architectural Adjustments
GraphRAG Iterative Retrieval Research 2026
Internal Architecture Repository
K-Complexity and Multi-File LLM Code Generation
LLM Grammar Constraints for Code
LLM Output Mediation and Programmatic Validator Generation
LLM-Native Language Design
Language Features Empirically Linked to LLM Code Generation Success
Local Autonomous Research Findings 2026
MENS Synthetic Corpus: Limitations and Mitigation Strategies (Research 2026)
Minimum Viable Corpus Size for QLoRA Domain Adaptation
Multi-repo context isolation: research findings 2026
Populi GPU network research 2026
Production Evidence: Context Truncation as a Silent Failure Mode
Production Failure Mode Catalog with Mitigations
Quality and Mode Collapse in Self-Play LLM Loops
Research Synthesis: Grand Strategy Seed 2026
Research: ASR Speech-to-Code Findings
Research: Claude Code Ultraplan Architecture
Research: Fuzzy & Partial Parsing
Research: Phonetic Operators vs. Symbols
Research: Planning Mode Capability Map
Research: Planning Mode and Agentic Coding 2026
Risk Taxonomy, Monitoring Design, and Open Research Questions
Scientia Publication Endpoints — Ground-Truth Research & Implementation Policy (April 2026)
State of the Art for Context-Aware Agent Handoff Protocols
Telemetry unification research findings 2026
Terminal AST validation research 2026
Terminal execution policy research findings 2026
The Compile-Pass Oracle and Semantic Degradation
The Efficacy of Binary Parse-Rate as a Primary Reward Signal
The Frontier: Unknowns in LLM-Native Language Design
The Optimization Landscape of Positive-Only Training Loops
The Risks of Agent-Generated Prose (Schola & Scientia)
Utilizing Parse Failures as Negative Examples
Vox Developer User Journeys: Intent vs. Actualization
Vox Language Testing Pipeline
Vox Scientia Gap Analysis (April 2026)
Vox VS Code Extension — Frontend Redesign Research (2026)
Vulnerabilities in AST-Based Coverage Scoring and Reward Hacking
Works Cited: AI Agent Context and Handoff
Works Cited: Continual Learning Flywheel Risks
Works Cited: GRPO Reward Shaping
Works Cited: Hallucination and Type-System Research
agent handoff contract 2026
cryptography research findings 2026
cryptography ssot 2026
orchestrator symphony research 2026
scientia external discovery research 2026
scientia pipeline ssot 2026
scientia socrates unification research 2026
SCIENTIA implementation wave playbook 2026
Scientia Community Publishing Playbook 2026
GUI, v0/islands, vision, and Mens Qwen — virtuous-cycle implementation plan (2026)
Orchestrator task attachment_manifest (RFC 2026)
MENS Corpus: Full Implementation Plan (2026)
Clavis Cloudless Implementation Catalog
Clavis Cloudless Threat Model V1
Context management implementation blueprint
Context management phase 1 backlog
MENS Research Track Blueprint 2026
Populi GPU mesh implementation plan 2026
Scientia Publication Pipeline — Full Implementation Plan v2 (2026)
Telemetry implementation backlog 2026
Telemetry implementation blueprint 2026
Telemetry retention and sensitivity SSOT
Telemetry taxonomy and contracts SSOT
Vox 0.4 Grand Migration Plan (Uncompressed)
Vox Agentic Loop Overhaul + MENS Syntax-Intelligence Blueprint
English-Core + Latin Alias Migration Ledger
vox-dei HITL Redirect
Contributors
Contributor hub
Documentation governance
Agent instruction architecture
Coding Agent Instructions
Continuation Prompt Engineering
CI & Quality
CLI baseline metrics
Documentation authority pointers
Command compliance SSOT
Doc inventory SSOT
Operations
Clavis Break-Glass Runbook
Clavis Cloudless Ops Runbook
VoxDB data cutover and telemetry sidecar runbook
Reference
ADR 001 — Burn Backend Selection for vox-tensor
ADR 003 — Native Rust Training Over Python
ADR 004: Codex over Arca over Turso
ADR 005: Socrates anti-hallucination SSOT
ADR 006: Mens full-graph Candle QLoRA with qlora-rs
ADR 007: qlora-rs multi-layer training API (Phase 2c architecture gate)
ADR 008: Mens transport
ADR 009: Hosted mens / BaaS (future scope)
ADR 010 — TanStack as the Vox web spine
ADR 011: Scientia publication manifest SSOT
ADR 012 — Internal Web IR strategy for Vox
ADR 013 — OpenClaw WS-first native interop
ADR 014: async-openai selective adoption (spike outcome)
ADR 015: Vox Docker/OCI portability SSOT
ADR 016: Oratio streaming Whisper and constrained decode
ADR 017: Populi lease-based authoritative remote execution
ADR 018: Populi GPU truth layering
ADR 019: Durable workflow journal contract v1
ADR 020: Populi mesh scaling — default transport posture
ADR 021: Generated workflow durability parity
ADR 023: Optional telemetry remote upload
Acceptance runbook — Mens HF fine-tune convergence
Agent Messaging & Orchestration Roadmap (Aspirational)
Architecture Decision Records (ADR)
Architecture Decision Records (index)
Automation primitives
Binary release artifact contract
Boilerplate metrics and KPI framework
CI runner contract
CLI command surface (generated)
CLI reference (redirect)
CLI scope policy
Changelog
Clavis SSOT
Codex / Arca compatibility boundaries
Codex BaaS scaffolding
Codex HTTP API
Codex legacy migration
Codex vNext — schema domains
Codex, Arca, and Rust import policy
Command compliance
Command surface duals (intentional)
Communication protocols
Compatibility and deprecation windows
Crate and build-lane migration map
Crate hardening matrix (rolling)
Crate topology buckets
Cross-platform Vox — runbook
Database Query Reference
Deployment: Docker, Compose, Coolify, CI (SSOT)
Deprecation policy — Mens native fine-tuning
Diagnostic taxonomy (compiler)
Direct turso:: usage allowlist
Doc inventory verifier (SSOT)
Docker image baselines (D05)
Environment variables (SSOT)
Environment variables (SSOT) (redirect)
Environment variables SSOT filename (redirect)
Explicitly out of scope for Rust migration
External repositories & workspace SSOT
Feasibility: full-graph Candle training (qlora-rs)
Forward-only migration charter
GitHub-hosted runner exceptions
HF fine-tune gap matrix (SSOT ↔ code)
HF fine-tuning capability matrix (code-grounded)
HIR legacy AST wrappers (inventory)
Hashing & Identity Builtins
Human-In-The-Loop & Doubt
Information-theoretic questioning protocol
Installation Reference
Language Syntax Reference
Language ergonomics principles
MCP HTTP gateway contract
MCP HTTP read-role governance contract
MCP tool registry contract
MENS curriculum — speech-to-code stages
MENS findings: Composer and Kimi (2026)
Mens / HF fine-tune — LLM PR checklist
Mens Architecture 2026 Synthesis
Mens Cloud GPU Training Strategy
Mens Coordination & Database Write Safety
Mens Coordination Workflow Guide
Mens LoRA / adapter ownership (vox-tensor vs vox-populi)
Mens external technology options
Mens laziness and accuracy audit
Mens local serving SSOT (Schola + orchestrator)
Mens measurement gap analysis
Mens native training SSOT (Candle QLoRA–first; Burn LoRA deprecated in dispatch)
Mens strategy inputs checklist
Mens train defaults (generated)
Mens training data (JSONL) contract
Mesh / Populi SSOT (CPU-first)
Migration metrics (script → vox ci)
Migration: backend-centric flags → fine-tune contract
Mobile and edge AI — SSOT
OpenClaw Discovery and Sidecar SSOT
Oratio & speech SSOT (Candle Whisper, no whisper.cpp)
Orchestrator bootstrap factory and daemon boundaries
Orphan surface inventory
Package management migration (2026)
Parser ambiguity and robustness inventory
Parser feature matrix
Phase 0 documentation baseline — signoff
Populi overlay personal cluster runbook
Populi remote execution rollout checklist
Populi work-type placement policy matrix
QLoRA Fine-tuning Data Strategy & SSoT
Reference: Decorator Registry
Reference: Type System
Repo reconstruction benchmark ladder
Research Notes: Achieving Serverless-like Performance with MCP
Review Anti-Pattern Catalog Contract
Review Fix Pairs Contract
Review Regression Challenges Contract
Rust ecosystem support contract
Rust pattern modernization — Wave 0 baseline
SCIENTIA SSOT handbook (glossary, vocabulary, checklists)
SCIENTIA publication automation SSOT
SCIENTIA publication readiness audit
SCIENTIA publication worthiness rules
Scientia publication failure playbook
Searching the Documentation
Socrates protocol — single source of truth
Speech capture architecture (edge vs backend)
Speech-to-code pipeline (Oratio → MCP → compiler → MENS)
Speech-to-code — operations, security, rollout
Standard Library Built-ins
Standard Library Reference
Standard library surfaces
Syntax K complexity telemetry (WebIR + emit)
TOESTUB self-healing architecture 2026
TanStack SSR with Axum (development topology)
TanStack web backlog
TanStack web roadmap
Tavily Integration SSOT
Telemetry and research_metrics contract
Testing Standard — SSOT
Trim, build, and defer (feature lifecycle)
TypeScript boundary policy
Unified orchestration — SSOT
VS Code extension and vox-mcp compatibility
Vox Documentation Style Guide
Vox Feature Builds & Capabilities
Vox IR Specification
Vox Skill Marketplace
Vox Web Architecture Analysis
Vox Webhook Integration
Vox database language surface (canonical)
Vox full-stack build artifacts — single source of truth
Vox full-stack web UI — single source of truth
Vox portability SSOT
Web Model Reference
Workflow enumeration (GitHub Actions)
Workspace root Cargo.toml (fix forward)
Zig-Inspired Deployment Architecture
cli
vox-cli build and feature inventory
MCP tool reference (legacy path)

Vox: The AI-Native Programming Language

"vox-dei HITL Redirect"

vox-dei HITL

[!WARNING] DEPRECATED The architecture for the vox-dei HITL crate is now documented in hitl-doubt-loop-ssot.md.