Faq

Last Updated: 2025-12-01

General Questions

1. What is SIL?

SIL (Semantic Infrastructure Lab) is a research lab building the Semantic Operating System - a 6-layer architecture that makes AI reasoning transparent, traceable, and composable.

Think of it like this: Linux provides an OS for computation. SIL provides an OS for meaning.

We're not building another LLM or agent framework. We're building the semantic substrate that makes intelligent systems interpretable and reliable.

Status: 11 projects, 5 in production, used daily.

2. How is SIL different from LangChain, AutoGPT, or other agent frameworks?

Key distinction: SIL is infrastructure, not a framework.

You could build LangChain ON TOP OF SIL. You wouldn't build SIL on top of LangChain.

Example:
- LangChain: "Connect this LLM to that vector database and chain these prompts"
- SIL: "Here's how to represent meaning persistently (Layer 0), transform it deterministically (Layer 4), and verify provenance (GenesisGraph)"

3. Is SIL production-ready?

Yes - 5 projects are in production:

1. reveal (v0.16.0 on PyPI) - Code exploration, 86% token reduction
2. morphogen - Cross-domain computation (audio + physics + circuits)
3. tiacad - Declarative parametric CAD in YAML
4. genesisgraph (v0.3.0) - Verifiable process provenance
5. sup - Computational notebook system

Production means:
- Available on PyPI or GitHub releases
- Used daily in real workflows
- Stable APIs with semantic versioning
- Comprehensive test suites

6 more projects are in alpha/research stages.

4. Who is Tia?

Tia is SIL's Chief Semantic Agent - a transparent, named AI agent who contributes to SIL development.

Important: Tia is not a person or co-founder. She is:
- A persistent semantic toolchain within the Semantic OS
- An agent that provides decomposition, pattern discovery, scaffolding
- A demonstration of how transparent agents extend human reasoning

Why name an agent?
- Transparency: If an agent contributes, that provenance is acknowledged
- Accountability: You know what work came from human vs agent reasoning
- Research: Demonstrates the "glass box, not black box" principle

The collaboration pattern:
- Scott (human): Judgment, taste, conceptual grounding, architectural constraints
- Tia (agent): Decomposition, pattern discovery, structural scaffolding, bandwidth
- Together: A single reasoning loop with every step visible

This is the future of work SIL is building toward: transparent human-agent collaboration.

5. Can I use SIL today?

Yes! Here's how:

Quick Start (5 minutes):

pip install reveal-cli
reveal your_code.py

Try Production Projects:
- reveal: Progressive code exploration
- morphogen: Cross-domain computation (examples)
- tiacad: Declarative CAD (tutorial)
- genesisgraph: Verifiable provenance (quickstart)

Explore the Ecosystem:
- Project Index - All 11 projects
- Tools Documentation - Production systems explained

Learn the Architecture:
- Quickstart - 30-minute guided tour
- Reading Guide - Choose your depth

6. What's the license?

Code: MIT License
Documentation: CC BY 4.0

In practice:
- ✅ Use SIL tools commercially
- ✅ Fork and modify projects
- ✅ Build proprietary systems on SIL substrate
- ✅ Cite and share documentation freely

Attribution appreciated but only required for documentation.

7. How mature is this?

It depends on what you're asking about:

Production-Ready (Mature):
- reveal (v0.16.0) - 2+ months in production, PyPI published
- morphogen (v0.11) - Used daily in cross-domain workflows
- tiacad (v3.1.1) - Declarative CAD, stable API
- genesisgraph (v0.3.0) - Provenance tracking

Research/Alpha (Early):
- Pantheon IR (Universal Semantic Intermediate Representation) - "Assembly language for meaning" enabling cross-domain transformations (Glossary)
- Agent Ether - Multi-agent coordination protocols (research stage)
- Semantic Memory - Persistent knowledge substrate (alpha)

Documentation (Comprehensive):
- Technical Charter - Formal specification complete
- Architecture guides - Unified framework documented
- Research papers - Semantic manifold transport, agent-help standard

Recommendation:
- Use production tools today - They're stable and valuable
- Watch research projects - Pantheon IR and Agent Ether are foundational but evolving
- Read the charter - Architecture is well-defined even if not fully implemented

8. How do I contribute?

Step 1: Understand SIL's Principles

Read SIL Principles (10 minutes). All contributions must follow these 5 design constraints:
1. Clarity - Explicit over implicit
2. Simplicity - Essential complexity only
3. Composability - Modules that combine predictably
4. Correctness - Formal verification where possible
5. Verifiability - Trace provenance and reasoning

Step 2: Pick a Project

Browse Project Index and choose based on your interests:
- Code exploration: reveal
- Cross-domain computation: morphogen
- Provenance: genesisgraph
- CAD/modeling: tiacad
- Formal representations: Pantheon IR (research)

Step 3: Check Project Guidelines

Each project has a CONTRIBUTING.md in its repository:
- reveal/CONTRIBUTING.md
- morphogen/CONTRIBUTING.md
- (Check individual repos for others)

Step 4: Start Small

• Look for "good first issue" labels
• Fix documentation typos
• Add test cases
• Implement small features

General Expectations:
- Write tests for all functionality
- Document design decisions
- Preserve semantic invariants
- Follow existing code style

Technical Questions

9. What is the Semantic Operating System?

The Semantic OS is a 6-layer architecture (Layer 0-5) for knowledge work:

Layer 5: Human Interfaces / SIM  ← CLIs, GUIs, agents you interact with
Layer 4: Deterministic Engines   ← Morphogen, hermetic builds
Layer 3: Agent Ether             ← Multi-agent coordination
Layer 2: Domain Modules          ← Water, Healthcare, CAD, etc.
Layer 1: Pantheon IR             ← Universal semantic types
Layer 0: Semantic Memory         ← Persistent knowledge graphs

Key Features:

1. Persistent Semantic Memory (Layer 0)
   - Knowledge that survives beyond single prompts
   - Graph-based with provenance tracking
   - Queryable, composable, verifiable

2. Universal IR (Layer 1)
   - Pantheon IR - "Assembly language for meaning"
   - Cross-domain interoperability
   - Types, operators, transformations

3. Domain Modules (Layer 2)
   - Water cycles, healthcare workflows, CAD geometries
   - Composable via shared IR
   - Domain-specific but semantically aligned

4. Multi-Agent Protocols (Layer 3)
   - Agent Ether - Coordination substrate
   - Transparent multi-agent collaboration
   - Inspectable reasoning chains

5. Deterministic Engines (Layer 4)
   - Morphogen - Cross-domain computation
   - Hermetic, reproducible execution
   - Formal verification where possible

6. Human Interfaces (Layer 5)
   - reveal, browserbridge, conversational agents
   - Every layer visible and inspectable
   - Progressive disclosure of complexity

Read more: Semantic OS Architecture

10. What is USIR / Pantheon IR?

USIR = Universal Semantic Intermediate Representation
Pantheon IR = SIL's implementation of USIR

Think of it as:
- LLVM IR for semantic computation (not just code)
- Assembly language for meaning
- Protocol for cross-domain interoperability

What It Provides:
- Types: Semantic primitives (entity, relation, transformation, constraint)
- Operators: Transformations with explicit semantics
- Composability: Operations that combine predictably

Example (conceptual):

# Instead of opaque strings:
"water cycle" → [string]

# Pantheon IR exposes structure:
WaterCycle(
  components=[Ocean, Atmosphere, Land],
  transformations=[
    Evaporation(input=Ocean, output=Atmosphere),
    Precipitation(input=Atmosphere, output=Land),
    Runoff(input=Land, output=Ocean)
  ],
  conservation_law=Mass(input) == Mass(output)
)

Why This Matters:
- CAD tools can understand water cycles (both involve flows and constraints)
- Physics simulations can verify water models (shared operators)
- Educational systems can explain water cycles (exposed structure)

Status: Research prototype, not yet production-ready.

Read more: Technical Charter, Section on USIR

11. How does SIL save $47K per year for agents?

Context: reveal's token reduction analysis (real production data).

The Math:

Before SIL (traditional RAG):
- Agent reads entire file to understand code: ~500 tokens/file
- 1000 agents × 100 files/day × 500 tokens = 50M tokens/day
- 50M tokens/day × $0.03/1K tokens = $1,500/day = $547K/year

After SIL (progressive disclosure):
- Agent sees structure first: ~50 tokens
- Only reads full function if needed: +150 tokens average
- 1000 agents × 100 files/day × 70 tokens = 7M tokens/day
- 7M tokens/day × $0.03/1K tokens = $210/day = $77K/year

Savings: $547K - $77K = $470K/year per 1000 agents
Or: $47K/year per 100 agents (the cited figure)

Key Insight:
- 86% token reduction isn't marketing - it's geometric
- Progressive disclosure (structure → details → full context) eliminates waste
- This pattern extends across ALL semantic infrastructure (not just code)

Real-World Impact:
- Organizations running 100+ agents save real money
- Faster responses (fewer tokens to process)
- Better results (agents see structure, not walls of text)

Read more: Tools Documentation, Economics section

12. What's the roadmap?

Year 1 (Research Agenda):
- Pantheon IR maturation - Universal semantic types stabilized
- Agent Ether protocols - Multi-agent coordination patterns
- Semantic Memory v1.0 - Production-ready knowledge persistence
- Cross-domain demos - Water+Healthcare+CAD integration

Near-Term (Next 6 months):
- reveal v0.14+ - Pattern libraries, more language support
- morphogen enhancements - Audio DSP, circuit simulation improvements
- GenesisGraph v0.4 - Compliance attestations, audit trails
- Documentation expansion - More use cases, tutorials, comparisons

Long-Term Vision:
- Glass box AI - Every reasoning step visible and verifiable
- Cross-domain composition - Tools that work together via shared semantics
- Civilization-scale infrastructure - The "steel" to AI's "wood"

Read more: Research Agenda Year 1

13. Can I see example code / demos?

Yes! Production examples:

reveal (Code Exploration):

pip install reveal-cli
reveal morphogen/src/core.py
reveal morphogen/src/core.py Operator

morphogen (Cross-Domain Computation):
- Audio synthesis examples
- Physics simulation examples
- Circuit design examples

tiacad (Declarative CAD):
- Full tutorial
- Example models

genesisgraph (Provenance):
- 5-minute quickstart
- Compliance examples

All project links: Project Index

14. How does SIL handle LLM non-determinism?

SIL's approach: Isolate and contain stochasticity.

Layer 4: Deterministic Engines
- Once semantic representations are formed, transformations are deterministic
- Morphogen executes workflows with reproducible results
- Hermetic build principles applied to semantic computation

Layer 5: Human Interfaces (where LLMs live)
- LLMs are input/output adapters (natural language ↔ semantic IR)
- Non-determinism is explicit and tracked
- Multiple generations can be compared at semantic level

Example: CAD Design

[User] "Create a bracket for mounting this sensor"
   ↓ (LLM translation - non-deterministic)
[Semantic IR] Bracket(constraints=[...], dimensions=[...])
   ↓ (Deterministic execution)
[CAD Model] STL file, always same for same IR input

Key Principle:
- Stochasticity at the edges (human interface)
- Determinism in the core (semantic computation)
- Provenance everywhere (track when and why randomness was involved)

Read more: Principles - Reproducibility principle

15. Where can I learn more?

Quick Paths:

30-Minute Overview:
→ Quickstart - Guided tour with hands-on example

Deep Architecture:
→ Unified Architecture Guide (20 min)
→ Technical Charter (45 min)

Philosophy & Vision:
→ Founder's Letter (10 min)
→ Manifesto (15 min)

Choose Your Path:
→ Reading Guide - 4 curated reading paths

Try Production Tools:
→ Tools Documentation
→ Project Index

Research Papers:
→ RAG as Semantic Manifold Transport
→ Agent-Help Standard

Community:
→ GitHub Organization
→ GitHub Issues on individual project repos

Still Have Questions?

For technical questions:
- Open an issue on the relevant project's GitHub repo
- Check project-specific documentation

For general inquiries:
- Email: (contact information coming soon)

For contribution questions:
- Read Contributing Guidelines
- Check project-specific CONTRIBUTING.md files

Created: 2025-12-01
Part of: SIL Documentation
License: CC BY 4.0