The 4-Phase Deterministic Architecture

Explore the machine-enforced lifecycle that isolates autonomous code generation and eliminates silent regressions.

Session Bootstrap & Subconscious Sync

nomos-handshake

Human PO triggers /nomos-handshake via chat. The Autonomous Agent automatically runs the Go substrate Doctor, asserts SQLite database integrity, and queries subconscious GitBrain memory & Git Notes.

🛡️ Machine-Enforced Invariants
  • Strict read-only lock on primary working tree
  • Automated SQLite health & migration validation
  • Subconscious GitBrain semantic vector memory loaded
Human Interaction & Substrate Automation
Human PO Chat
/nomos-handshake
AI Agent Substrate
$ nomos handshake
Key Objective: Initializes workspace state, verifies toolchains, and injects cognitive memory.
⚡ 3-Step Zero-Risk Pilot Human PO ⇄ AI Agent Symbiosis

Try One Task on Your Existing Repo

Experience deterministic AI engineering rails without changing how your team builds. You speak in natural language and slash commands; your AI agent operates the compiled Go engine underneath.

01 ONE-TIME SETUP

Install & Handshake

Run the 1-liner to install the Go binary, then type /nomos-handshake in your AI chat (Cursor, Claude Code, Antigravity). Syncs workflows into your IDE and arms pre-commit hooks.

$ curl -fsSL https://sophialabs.ca/install | sh
chat > /nomos-handshake
02 TRANSIENT SANDBOX

Command via /nomos-build

Describe your task in chat: /nomos-build "Fix issue #42". Your agent mounts an isolated Git worktree, runs the TDD loop, and passes 39 AST gates while main stays protected.

chat > /nomos-build "Fix issue #42"
agent > nomos task start → worktrees/repo-42
03 ATOMIC MERGE

Review & Fast-Forward

Review the agent's structured walkthrough diff in chat and type approve. Your agent executes atomic release sync, fast-forward merges to root, and tears down the sandbox.

chat > approve
agent > nomos task sync --approve

Beyond Prompt-Based AI: Deterministic Rails vs. Stochastic Prompts

Why prompting an LLM to "follow rules" fails in production, and how compiled binary tooling enforces guaranteed architectural boundaries.

Engineering DimensionPrompt-Based AI (Claude Code, OpenCode, Aider)Nomos & SophiaLabs (Deterministic Rails)
Enforcement Substrate How rules are appliedStochastic Prompt Suggestions (Honor System): The agent is given markdown instructions ("Please run tests", "Please don't edit main"). If the context window degrades, the model skips them without notice.Compiled Go Substrate & OS Hooks: Git pre-commit hooks and machine-enforced gates physically block forbidden mutations. If nomos verify fails, code cannot be checked in.
Working Tree Isolation Multi-agent safetyLive "Dirty" Working Tree: Agents edit files directly inside your active checkout, colliding with human edits, dirtying branches, and breaking running local dev servers.Transient Sandboxed Git Worktrees: Every task mounts an isolated directory (.nomos/worktrees/NOM-123) auto-linked via go.work/workspaces. Parallel agents execute in complete isolation.
Phase Discipline Plan vs. Execution lockUnrestricted Arbitrary Writes: The agent can jump from planning straight into refactoring or deleting source files on a whim without human sign-off.Cryptographic Phase State Locks: Repository files are locked read-only in PLAN. No source file can be touched until a valid phase transition token is issued by the human PO.
Definition of Done (DoD) Quality & Regression GatesLLM Self-Evaluation ("Vibe Check"): The model evaluates its own code in English text ("Looks good, all requirements met!"), easily ignoring inverted tests or dead code.38 Strict Binary Verification Gates: Zero-tolerance Go-native AST analysis inspecting cyclomatic complexity (< 15), docstring density, goroutine lifecycles, and Inner-Loop TDD assertions.
Inference Efficiency Token spend & model viabilityContext Thrashing & Retry Tax: Blind edits lead to broken builds. The agent re-reads 100k+ token contexts across 10-turn retry loops ($2.50+ per feature) and forces reliance on expensive frontier cloud APIs.85%+ Lower Token Spend & $0 Local SLMs: Sandboxed worktrees keep contexts under 8k tokens. 12ms AST line diagnostics resolve errors in 1 turn, unlocking production viability for fast, private local models.
The 90-Day Velocity Curve Speed vs. Long-Term HealthDay 1 Speed ➔ Day 90 Sludge: 10x code output feels fast initially, but unconstrained drift results in 80% of human time spent untangling AI spaghetti, dead imports, and phantom bugs.Compound Acceleration ("Slow Down to Go Fast"): Spending 15 seconds executing sandboxed TDD and AST gates ensures code is written once, verified immediately, and never regressed.

"The Cognitive Layer explores, designs, and refactors; the Substrate Kernel enforces, verifies, and guarantees. We never rely on an LLM to 'remember' constraints if those constraints can be compiled into deterministic verification gates."

The Core Axiom — Harness Over Model
Read Full Essay →
⚡ Interactive Verification Simulator

Stochastic Chaos vs. Deterministic Go Rails

Drag the slider to test how unmanaged prompt loops fail at scale versus physical AST-enforced compiler gates.

🔒 39 DoD AST Gates Locked

Nomos Go Deterministic Substrate

Zero drift, instant 12ms AST verification, 100% transient worktrees linked via go.work.

0% Vibe Drift 50% Prompts 100% Nomos Go
Verification Strictness: 100% [PASS]
AST Cyclomatic Complexity
6 (< 15 Target)
Test Suite Regression Status
✓ 100% PASS
Token Burn / Turn
$0.00 (Local SLM)
Git Workspace State
Isolated .nomos/worktrees/SOP-4
AST Compiler Sandbox Continuous Verification

Real-Time AST Definition of Done Sandbox

Type or paste code to test cyclomatic complexity, docstrings, security invariants, and empty blocks in sub-10ms.

22 lines • 725 chars
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22

Why Engineering Teams Choose Nomos

Moving from unanchored prompt-based AI coding to machine-verified deterministic software engineering.

The Status Quo: Unanchored AI Agents

Stochastic Vibe Coding & Silent Regressions

When AI coding assistants write code without deterministic verification, codebases degrade into unmaintainable bloat:

  • Direct Protected Commits: Agents commit half-finished experiments directly to master or main.
  • Interleaved Multi-Agent Collisions: Parallel sessions run git add . and corrupt each other's staged files.
  • Tautological Tests: Agents write tests after implementation, silently testing inverted logic.
  • Runaway Complexity: Functions balloon with unvetted nested branches and orphaned dead code.
The Nomos Paradigm: Machine-Enforced Invariants

Autonomous Engineering OS & 36 Definition of Done Gates

Nomos acts as a rigorous Go harness that governs all autonomous agent actions inside machine-enforced rails:

  • Transient Worktree Isolation: Every task executes in an isolated sandbox; zero branch pollution.
  • Mandatory Test-First TDD: Agents must write failing tests before touching application logic.
  • Automated AST Quality Audits: Cyclomatic complexity (< 15), docstrings, and imports checked on every change.
  • Human-in-the-Loop Release Gates: Structured walkthroughs signed before fast-forward merges.

Built for Data Sovereignty, Low Inference Spend & Institutional Rigor

Designed from first principles for on-premises security, 85%+ lower token burn, and pristine codebases.

📉

85%+ Lower Inference Spend

Transient worktrees and 12ms Go AST diagnostics eliminate context thrashing and 10-turn retry loops, slashing token burn and making $0 local SLMs viable in production.

🔒

100% On-Premises & Local Inference

Compatible with local Small Language Models (SLMs) via Ollama and vLLM. Your source code and architectural knowledge never leave your infrastructure.

Compiled Go Core Engine

Sub-millisecond static AST parsing, SQLite transactional state, and instant pre-commit hooks ensure zero IDE latency.

📦

Universal Language Support

Pre-configured verification profiles for Go, TypeScript/SvelteKit, Python, and Rust with native lint and test runners.

Sophia AI • Live Architecture Chat