The Admiral's Bridge

Where one engineer + AI orchestrates a thousand-ship fleet

AdmiralsBridge is the command center for StreamRift's entire operation. Built on HM0 — the M3 Ultra with 96GB — it is the portal between the Admiral and everything his empire touches. Not just battle-ready, but beautiful. Not just functional, but transferable. This is the seat that will one day be handed to the next in command.

96GB
Unified Memory
4+7
Compute + WHM Nodes
1,000
Target Projects
Machine HM0 // M3 Ultra
Date 2026-02-27
Status Genesis
Operator StreamRift // T4 Admiral

The Vision

In 2020, StreamRift imagined a framework that could launch 1,000 companies — so fathers could provide for their families and disabled women wouldn't live in a society that treats them as rodents. That vision wasn't a business plan. It was a conviction that building is the most moral act an engineer can perform.

"The profound flex of all engineering isn't just doing the thing, but doing it elegant and stylishly. You have bandwidth for the nice touches. It's not just bending time on what is possible, but it's inspiring just to engage with it."

Project 1,000

Not a thousand half-starts. A thousand real, shipped systems. Each one pushes into a domain not yet mastered, solves a problem that requires learning something new, and feeds knowledge back into the portfolio. The portfolio is the product.

Trading

G1XL

Real-time prediction market terminal for high-frequency sports bettors with $50-100k bankrolls. Live at kt.gbuff.com.

Hosting

BigTB.ai

Productized infrastructure knowledge into a commercial hosting offering on sovereign American hardware.

Creative

NewBx / NewBx Lab

Fearless experimentation sandbox where discoveries inform production systems.

Orchestration

OrchestrationFramework

Coordinating AI agents at scale across the entire portfolio.

The Operator Model

One person with AI as force multiplier. The COO model isn't a gimmick — it's the architecture that makes it possible for one engineer to run a portfolio that would normally require a team. Structure beats intelligence. Session continuity is everything.

Why the Bridge

NX-03 is a roaming piece of hardware. It's fun, it's mobile, but nothing about it undoes the fact that it's not a base station. HM0 — the M3 Ultra with 96 gigs — is the Admiral's office. This is where we observe and execute. The portal between StreamRift and everything his empire touches.

And at some point, this seat transfers to Brendan. It needs to be so well-architected that the transfer is a key turn, not a knowledge dump.

Fleet Command

HM0 is not just the strongest machine — it is the only machine that should hold the unified state of the entire operation. Star topology, SSH transport, JSON state.

THE FLEET ┌────────────────────────────────────────────────┐ │ ADMIRAL'S BRIDGE │ │ HM0 (M3 Ultra, 96GB) │ │ │ │ fleet-ctl ── fleet-state.json ── fleet-log/ │ └──────┬──────────┬──────────┬──────────┬─────────┘ │ │ │ │ ┌─────┴───┐ ┌───┴────┐ ┌──┴───┐ ┌───┴──────────┐ │ NX-03 │ │ NX-04 │ │ H0 │ │ WHM (1..7) │ │ M4 Air │ │ 5080 │ │M4 Pro│ │ Web Fleet │ │ 24GB │ │ 16GB │ │ 36GB │ │ │ │ Edge │ │ GPU │ │ Mid │ │ Hosting │ └─────────┘ └────────┘ └──────┘ └──────────────┘

Design Principles

  • SSH is the transport layer. No custom daemons on workers. No new network protocols. SSH multiplexing for performance.
  • JSON over stdout. Every fleet command returns structured JSON. Human-readable output is a presentation layer on top.
  • HM0 is the single source of truth. Other machines are workers; they report, they do not coordinate.
  • State file, not database. fleet-state.json is a flat file. For 11 nodes, this is correct. Revisit at 50+.
  • Zero deployment burden on workers. Workers run Ollama and accept SSH. All coordination logic lives on HM0.

fleet-ctl — The CLI Orchestrator

fleet-ctl status                  # Full fleet health snapshot
fleet-ctl ping                    # Quick heartbeat (<2s, parallel)
fleet-ctl dispatch <node> <cmd>   # Run command on remote node
fleet-ctl deploy <project> <node> # Push project to target
fleet-ctl sync <project>          # Git sync across mirrors
fleet-ctl logs [node] [--tail]    # Aggregated or per-node logs
fleet-ctl bench <node> <model>    # Inference benchmark
fleet-ctl ollama <node> <action>  # Proxy Ollama commands

Heartbeat System

A launchd plist runs fleet-ctl ping every 60 seconds. Checks SSH reachability, Ollama status, system metrics, and WHM health. Writes to fleet-state.json. Alerts drop to .hand/mail/ for Forge visibility on next boot.

Integration with /run-it

The boot sequence gains Step 0: read fleet-state.json, report fleet health before project status. Say "run it" and Forge reports the fleet alongside everything else.

The Dashboard

An operational cockpit, not a monitoring dashboard. The single browser tab StreamRift opens when he sits at HM0. Answers "what's happening across my entire operation?" in under 15 seconds.

Nine Panels

Core

Fleet Overview

All machines, health, load, current tasks, inference status.

Core

Portfolio

All projects, status, last activity, lifecycle stage, health.

Core

Sessions

Active Claude Code sessions across machines. Quick-launch.

Ops

Inference Manager

Model loading, fleet-wide inference allocation, benchmarks.

Ops

Pipelines

Build/deploy pipelines, Smithy status, CI results.

Ops

Hosting

WHM servers, accounts, disk, bandwidth, site health.

Infra

Storage

Disk usage across fleet, backup status, model storage.

Phase 2

Arena

Model benchmarking arena, head-to-head comparisons.

Phase 3

War Room

Strategic planning view, decision tracking, roadmaps.

Stack

React 18 + Vite + Tailwind + Express + WebSocket. Zero new technology — identical to G1XL and C2. Absorbs existing C2 fleet monitoring and Launchpad project discovery into a unified command surface.

Aesthetic: Naval Command, Not Grafana

Deep navy-dark (#0a0e14), teal primary accent, monospace data, flat panels with 1px borders, 150ms transitions. No decoration without function. Information-dense but not cluttered. The name "Admiral's Bridge" sets the tone.

Agent Architecture

Two Forges. A federated MCP mesh. Personas decoupled from models. This is how Forge becomes Fleet Admiral — not just one Claude session on one project, but the orchestration layer across the entire fleet.

The Two Forges

Admiral

Forge Admiral (HM0)

Portfolio-wide coordinator. Dispatches work across the fleet. Holds fleet context. Lives in AdmiralsBridge repo.

Roaming

Forge Roaming (Any Machine)

Project-scoped executor. Writes code, runs tests, ships. Lives in any project repo. Reports to Admiral via mail.

The MCP Mesh

mcp-admiral on HM0 is the single MCP tool surface that Claude Code calls. Admiral routes to satellite HTTP services on NX-03, NX-04, and H0. Satellites operate independently if Admiral is down. Claude never talks to satellites directly.

Agent Personas

Gary Jr, Quinn Jr, Oscar — these are stable identities with defined capabilities. Whether Gary Jr runs gemma3:1b or gemma4:2b is a YAML config swap. Model upgrades never break dispatch logic. Personas are the contract, models are the implementation.

Context Hierarchy

LayerContentLoaded When
L0Portfolio summaries (~100 lines)Admiral boot
L1Project-specific contextOn demand (hail)
L2Source code, deep analysisDeep dive
L3Mechanical workDelegated, never enters Admiral

Scaling to 1,000

At 100+ projects, category-level Sub-Admirals (Trading Admiral, Infra Admiral, Creative Admiral) sit between Forge Admiral and project-level Roaming instances. Each is a Claude session with category-scoped context.

The Nerve Center

HM0 audited live. 96GB, Ollama running 22 models, Postgres with live trading data, 14 SSH keys. Critical gaps: no backups, no monitoring, fleet nodes going offline undetected.

What's Running on HM0 Today

22
Ollama Models
373GB
Model Storage
15
Postgres Tables

Port Allocation

PortServiceStatus
11434OllamaRunning
5432PostgreSQLRunning
3100MCP AdmiralPlanned
3200Fleet GatewayPlanned
5300Bridge DashboardPlanned
6379RedisPlanned

Resource Budget (96GB)

AllocationGBPurpose
Inference60Model loading, embeddings, on-demand heavy models
Development16IDE, Claude Code, build tools, browsers
OS + System12macOS, system processes
Services3Postgres, Redis, MCP, fleet services
Reserve5Headroom for spikes

Fleet Memory

The context window is where you think; the archive is what you know. Three-tier memory turns "AI that helps with one project" into "genuine second brain for the portfolio."

Three Tiers

Warm // Per-Project

Tier 1: .hand/

MEMORY.md, reports, directives, ADRs. Loaded on boot. Updated on wrap. What exists today, made useful.

Fleet // Portfolio

Tier 2: Fleet Index

Session log index (all sessions, all projects), decision registry, pattern library, operator memory. YAML on HM0.

Cold // Semantic

Tier 3: Vector Search

SQLite + nomic-embed-text embeddings + sqlite-vec. Semantic search across all reports, ADRs, patterns. ~400MB idle.

Compression Pipeline

A 2-hour session (50K tokens) compresses through four stages:

StageOutputSize
Session transcript~50K tokensFull conversation
Report~500 linesStructured debrief
MEMORY.md entry~5 linesHot state for next boot
Session log entry~8 linesIndexed for fleet search

Compression ratio: ~1000:1. This is how 1,000 projects stay manageable.

The fleet_recall MCP Tool

Natural language queries across the fleet's memory. "Why did we choose X for project Y 6 months ago?" returns ranked results in under 100ms. Powered by local embeddings on HM0 — no API calls, no data leaves the machine.

The Handoff

The difference between "here's my password" and "here's a system designed for succession" is that the password gives access, but the system gives competence. The operator becomes a variable in the system, not a hardcoded constant.

Operator Profiles

.hand/ACTIVE_OPERATOR contains one word. The boot sequence loads the corresponding profile from .hand/operators/. Switching operators is a deliberate act, not a guess.

Permission Tiers

TierNameWhat They Can Do
T1ObserverRead-only. /sitrep, /scout. Bridge explains everything.
T2ApprenticeCommit to dev. /forge-mode, /crucible. No production deploy.
T3OperatorFull project execution. Deploy, ADRs, directives.
T4AdmiralFull authority. Can modify the Bridge itself.

COO Adaptation

Forge adjusts across six axes per operator:

StreamRift (T4)

Admiral Profile

Terse explanations. High autonomy. Light guardrails. Sandwich format. Full portfolio access.

Brendan (T1)

Observer Profile

Thorough explanations. Low autonomy. Heavy guardrails. Tutorial format. Current-project only.

/bridge-training

Interactive 4-phase onboarding skill. Phase 1: Orientation (interview, tour, comprehension check). Phase 2: Supervised practice. Phase 3: Independent operation. Phase 4: Full transfer. The Bridge teaches itself to the new operator.

Project 1,000

Six interconnected systems that take the portfolio from 20 projects to 1,000. File-based until files don't scale. Commands, not UI.

The Six Systems

Foundation

Project Registry

portfolio.json — single source of truth. Every project's name, stage, tech stack, dependencies, health. ~500KB at 1,000 projects.

Creation

Scaffolding Engine

/new-project — zero-to-running in one command. Template catalog for web apps, CLI tools, API services, infrastructure.

Visibility

Portfolio Dashboard

/bridge — 30-second health view. Exception-based at 100+. Category rollup at 1,000.

Integrity

Dependency Graph

graph.json — what breaks when something changes. Cross-project dependency tracking.

Intelligence

Knowledge Mesh

/learn — cross-project learning. Patterns graduate when adopted across 3+ projects, then bake into scaffold templates.

Focus

Attention Allocator

/focus — manages the scarcest resource. Weekly allocation. Scoring prevents critical dependencies from rotting.

Lifecycle Stages

idea → scaffolding → active → maintenance → stable → archived → sunset

At 1,000 projects, only 5-15 are in active development at any time. The Bridge surfaces the right tooling for each stage. The rest are either running autonomously or waiting.

Local Inference

96GB unified memory unlocks the Specialist+ tier. Models that don't fit on any other machine in the fleet run here. 60-70% of Claude API work can stay local, sovereign, free.

The Always-On Trio (~32GB)

ModelParamsMemorySpeedRole
qwen2.5-coder:32b32B~23GB25-35 tok/sCode generation workhorse
gemma3:12b12B~8GB50-80 tok/sFast utility, summarization
mxbai-embed-large137M~1.2GBN/AEmbeddings for fleet RAG

On-Demand Heavy Models (~40-43GB)

ModelParamsMemorySpeedUse Case
deepseek-r1:70b70B~43GB12-18 tok/sChain-of-thought reasoning
llama3.3:70b70B~40GB12-18 tok/sGeneral intelligence
qwen3-coder:30b-a3b30B MoE~18-20GB40-68 tok/sFast MoE coding
minicpm-v:8b8B~5-6GB30-50 tok/sVision / OCR

Economics

$47-110
Estimated Monthly API Savings
60-70%
Tasks Routable to Local

Tier System (Updated)

TierModelsSpeedWhere
Nanogemma3:1b, phi-3-mini89+ tok/sNX-03, any
Workergemma3:4-8b, qwen3:1.7-8b10-57 tok/sNX-03, H0
Specialistcodestral, deepseek-coder:33b5-15 tok/sH0, HM0
Specialist+qwen2.5-coder:32b, 70B models12-35 tok/sHM0 only
ArchitectClaude Opus/SonnetAPICloud

Security & Sovereignty

The audit found the Bridge running with shields down. FileVault off. Firewall open. Zero backups. This is the foundation that must be laid before anything else.

Critical Finding: If HM0's SSD fails right now, 14 SSH keys, the streamrift trading database, and 373GB of Ollama models vanish. FileVault is disabled — a stolen Mac means full access to everything.

Seven Security Layers

#LayerStatusImplementation
1Disk EncryptionNot enabledFileVault, recovery key in safe
2Network HardeningWide openFirewall + stealth, SSH key-only, Ollama localhost
3Secret ManagementNo systemage + sops, encrypted YAML, pre-commit hooks
4Overlay NetworkFlat LANTailscale mesh, per-role ACLs
5Backup & DRNothingrestic + Backblaze B2 (daily), local weekly
6Data SovereigntyImplicitClassification (CRITICAL/HIGH/MEDIUM), rules
7Audit & HygieneMinimalSSH logging, GPG commits, update schedule

Day 0 Hardening (Under 1 Hour)

Runbook at .hand/runbooks/security-hardening-day0.md. Five steps: FileVault, macOS firewall + stealth mode, SSH key-only auth, fleet pf rules, Ollama localhost binding. Each step has verification commands and rollback procedures.

The Portal Experience

Beyond the architecture — what it feels like to sit at the Bridge. The boot-up ritual. The 30-second awareness cascade. The command vocabulary. Engineering as art applied to the meta-tool itself.

The Boot-Up Ritual

0-3 seconds

Awareness

Terminal opens. PS1 prompt shows machine identity, branch, fleet count. Tmux layout materializes.

3-10 seconds

Sweep

Background fleet health check. SSH pings, Ollama status, alert collection. Cached before you ask.

First command

Briefing

Data already cached. "run it" produces fleet + project + recommendation instantly. You never wait.

The Terminal Layout

┌────────────────────────────────────┬──────────────┐ │ │ FLEET │ │ MAIN (80%) │ (10%) │ │ │ │ │ Claude Code / Forge session │ hm0 ● ok │ │ The sacred workspace │ nx03 ● ok │ │ │ nx04 ○ off │ │ │ h0 ● ok │ │ ├──────────────┤ │ │ WATCH │ │ │ (10%) │ │ │ │ │ │ event tail │ │ │ log stream │ └────────────────────────────────────┴──────────────┘ ┌───────────────────────────────────────────────────┐ │ hm0:admirals-bridge (main) [a3f2c1d] +2 ~1 │ F:4 │ └───────────────────────────────────────────────────┘

Command Vocabulary (Extended)

Smithy

Existing Commands

run it, sitrep, forge, crucible, anvil, ship it, hammer time, wrap it, make it so, scout, architect, war room

Fleet

Bridge Extensions

hail <project> — switch focus. fleet status — full report. battle stations — focus mode. shore leave — graceful shutdown.

The Three Promises

1. You will never start cold. The Bridge is always aware before you ask.
2. You will never fight the process. The Smithy handles workflow. You handle creation.
3. You will never outgrow the tool. Twenty projects or a thousand. The limitation is always external.

The Armory

The gear room. Head down to see how much firepower the fleet has and how to deploy it. 22 models, 373GB on disk, 96GB of unified memory, and the question: how many parallel Forge instances can we run?

Live Benchmarks (HM0, Feb 27 2026)

ModelParamsStorageGen tok/sTierVerdict
gemma3:1b1B0.8GB177-191NanoScreaming fast. Nano king.
gemma3:4b4B3.1GB100-112Nano+Best speed:quality ratio
gpt-oss:20b21B12.8GB99-102Worker*MXFP4 miracle. Best worker.
nemotron-3-nano31.6B MoE22.6GB46-55Worker+MoE speed, huge knowledge
qwen3:8b8B4.9GB47-52WorkerSolid, reliable
glm-4.7-flash29.9B MoE17.7GB14SpecialistVerbose, slow for size
qwen2.5-coder:32b32.8B18.5GB9.8-11.5SpecialistBest code quality, slow
qwen3-coder:30b30.5B MoE17.3GB8.6SpecialistDisappointing speed

The gpt-oss:20b revelation: MXFP4 quantization pushes 21B parameters at 102 tok/s — nearly matching gemma3:4b speed with dramatically better quality. This is the fleet's best worker model.

The Breakthrough: OLLAMA_NUM_PARALLEL

You do NOT need to load a model N times to serve N agents. Ollama's num_parallel lets one loaded model serve multiple concurrent requests. The RAM cost for "6 agents using qwen3:8b" is ~7.9GB (model + 6x KV cache), not 29.4GB. HM0 is running with factory defaults — that single config change unlocks everything.

Named Loadout Profiles

8 Agents // 3.2GB

Nano Swarm

gemma3:1b p=8. Maximum parallelism for mechanical tasks. File ops, templates, import updates.

6 Agents // 15.4GB

Speed Squad

gemma3:1b p=4 + gpt-oss:20b p=2. Fast nano workers + quality mid-tier.

6 Agents // 22.5GB

Balanced Ops

gpt-oss:20b p=2 + qwen3:8b p=2 + gemma3:1b p=2. Recommended default.

6 Agents // 24GB

Code Factory

qwen2.5-coder:32b p=2 + gemma3:1b p=4. Quality code gen + fast workers.

2 Agents // 49.7GB

Heavy Hitter

llama3.3:70b p=1 + gemma3:1b p=1. Maximum reasoning + a runner.

1 Agent // 65GB

Titan Mode

gpt-oss:120b p=1. The 117B behemoth. Full power, single focus.

Ollama Config (Apply Today)

launchctl setenv OLLAMA_MAX_LOADED_MODELS 4
launchctl setenv OLLAMA_NUM_PARALLEL 4
launchctl setenv OLLAMA_KEEP_ALIVE "-1"
launchctl setenv OLLAMA_FLASH_ATTENTION 1
# Then restart Ollama

Fleet Capacity (All Nodes Online)

172GB
Total Fleet Memory
~21
Max Agent Slots
14
Unbenchmarked Models

Benchmark Framework

Six suites. 24 test prompts. Automated scoring against regex patterns. Speed + quality combined into a single fitness score, weighted by task type.

SuiteTestsQuality WeightMeasures
CODE_GEN570%Write correct code from spec
CODE_REVIEW485%Find real bugs, not noise
FILE_OPS360%Mechanical precision
SUMMARIZE375%Extract facts, don't hallucinate
REASONING390%Step-by-step logic
CREATIVE365%Structured professional prose

After ollama pull, the auto-benchmark trigger runs all suites immediately. Historical results stored in JSONL. Leaderboard generated on demand. Fleet-wide comparison shows same model across different hardware.

Implementation Roadmap

From empty repo to full fleet command in 10 weeks. Security first. Then eyes. Then everything else.

P0 — Today: Run the Day 0 security hardening runbook. FileVault + firewall + SSH hardening + Ollama binding. Under 1 hour. Everything else builds on this foundation.

Phase Timeline

P0
Day 0

Security Hardening

FileVault, firewall, SSH key-only, Ollama localhost, Time Machine. Under 1 hour.

P1
Weeks 1-2

Fleet Visibility

fleet-ctl CLI, node registry, SSH multiplexing, heartbeat daemon, fleet-state.json, /run-it fleet integration.

P2
Weeks 3-4

Fleet Dispatch & Deploy

Remote execution, cross-machine deployment, git sync, post-Anvil fleet sync hook.

P3
Weeks 5-6

MCP Bridge & Inference

Ollama MCP server, fleet-aware routing, inference dispatch, always-on model trio.

P4
Weeks 7-8

Memory & Knowledge

Portfolio memory index, session log, local embeddings, SQLite-vec, fleet_recall MCP tool.

P5
Weeks 9-10

Dashboard v1

React cockpit absorbing C2, fleet + portfolio + sessions panels. The visual Bridge.

What Ships When

MilestoneCapabilityWeek
"run it" shows fleetFleet health in every boot2
Dispatch from BridgeRun commands on any machine4
Local inference routingForge delegates to fleet models6
Semantic fleet memory"Why did we do X?" answered in 100ms8
Visual cockpitFull dashboard in browser10
Brendan onboarding/bridge-training ready10

The Decision

Ten ForgeStorm sessions. Six architectural layers. Eighteen documents. One empty repo that is now a blueprint for the seat of an empire.

The Admiral's Bridge is where one engineer + AI punches so far out of his weight class that he surprises even himself. Not just because of what it can do — but because of how it feels to sit in that chair.

The build starts with P0. One hour. Shields up.