Theme
Field Manual · Edition 90 · Own Your Compute · No Cloud Landlord Required

CLASSIFICATION: UNCLASSIFIED // DISTRIBUTION: YOURS ALONE

Personal
Data Center

Improvise. Adapt. Self-host.

The sovereign way to run software: rig the exact tool you need from parts you own. A Start9 server, open-source services, your own AI — wired together and reachable anywhere. No cloud landlord. No vendor in the loop.

Read the Brief Open OP 01

START9 SERVER — THE BOX NO CLOUD LANDLORD PWA — OPEN ANYWHERE STARTOS — RUNS YOUR SERVICES START-TUNNEL — HOME IP HIDDEN

Bayside High · Computer Lab · Est. wherever your server lives

Saved by
the PDCPersonal Data Center

Your compute. Your keys. Your off switch.

The sovereign way to run software. Build the app you actually need, run it as a service on your own Start9 server, and open it anywhere as a progressive web app. No platform standing between you and your compute.

The Build Log

// four operations — and the four scenes we shoot

+ APPENDIX A — TOOLS OF THE TRADE →

Today's Bell Schedule

// also the shooting schedule — one period per scene

★ Extra Credit — Tools of the Trade →

Period 1 — Homeroom

The Sovereign Stack

A Personal Data Center is the hardware and software you own and run yourself. Instead of renting compute from someone else's cloud, the box sits where you live. You hold the keys, the data, and the off switch.

OP 01

// HOMEROOM — THE SOVEREIGN STACK

Build It From Parts You Own

A Personal Data Center is the hardware and software you own and run yourself. Instead of renting compute from someone else's cloud, the box sits where you live. You hold the keys, the data, and the off switch.

  • The box. A Start9 server is the hardware — a small machine you own outright.
  • The OS. StartOS packages and runs software as services. Install one the way you'd install an app, except it runs on your hardware.
  • Bespoke software. Build the exact app you need, package it as a StartOS service (.s9pk), and install it next to everything else.
  • PWA access. Open the service on your phone or laptop, pin it to the home screen — native feel, your box behind it.
  • Reach, privately. Keep it local, or publish a single service to the open web through Start-Tunnel — a VPN-style tunnel that keeps your home IP off the grid.
startos // field readout
# build the app you actually need,
# package it as a service, install it on your box
$ start-cli package build ./the-daily-vol
✓ built the-daily-vol.s9pk
$ start-cli package install the-daily-vol.s9pk
✓ service running
$ start-cli net tunnel enable the-daily-vol
✓ reachable on the open web — your call
# traffic routes through a VPS; visitors see
# the tunnel's IP, never your home address

LEAVE NO TRACE

Start-Tunnel runs the connection out through a VPS, so the address the world sees is the tunnel's — not your house. Like a VPN for your services: publish only what you choose to the open web, and your home connection stays off the grid.

// shielded by Start-Tunnel — VPS out front, home server behind it

Hall Pass: your home IP stays off the record

Start-Tunnel runs the connection out through a VPS, so the address the world sees is the tunnel's — not your house. Like a VPN for your services: publish only what you choose to the open web, and your home connection stays hidden.

// shielded by Start-Tunnel — VPS out front, home server behind it
Period 2 — A.I. Lab

Building with AI

Frontier models do the heavy lifting on the first build, the security audit, and the evals. You stay the editor-in-chief: the model writes, you review, test, and decide what ships.

OP 02

// FIELD LAB — BUILDING WITH AI

Let AI Do the Heavy Lifting

Frontier models do the heavy lifting on the first build, the security audit, and the evals. You stay the editor-in-chief: the model writes, you review, test, and decide what ships.

  • Heavy lifting up front. Describe the system; a frontier model drafts the first version, audits it for security, and helps write the evaluations.
  • OpenClaw is the harness. It turns a spec into commits — one concern per commit, type hints and Pydantic throughout, nothing you didn't ask for.
  • Patch from many models. Hit a wall, route the fix to several APIs — Opus, Grok, Venice — and keep the patch that holds under your tests.
  • You hold the merge. The model proposes; you read the diff, run the evals, and decide. No black box gets commit rights.
openclaw // patch the bug
$ openclaw issue "feed ingest dies on bad RSS date"
# fan the fix out to several models
$ openclaw patch --models opus,grok,venice
opus    ✓ passes evals
grok    ✓ passes evals
venice  ✗ regression in parser
$ openclaw commit --pick opus
✓ one concern, one commit → GitHub
spec → commitsecurity auditevals before mergeyou review the diff
Period 3 — Computer Class

Your Own Weights

The services that run your PDC are open source — read them, fork them, host them. And the model itself can be yours: fine-tune an open model on your own data, on your own GPUs, with nothing leaving the building.

OP 03

// THE WORKBENCH — OPEN SOURCE ON YOUR BOX

Your Own Weights

The services that run your PDC are open source — read them, fork them, host them. And the model itself can be yours: fine-tune an open model on your own data, on your own GPUs, with nothing leaving the building.

  • Open source, on your box. No license server, no phone-home. The code runs because you run it.
  • LLaMA-Factory. A fine-tuning toolkit you run locally — point it at your data and your hardware.
  • Fine-tune Gemma 4. Train it on your own corpus — say, your own writing — to get a model that answers in your voice.
  • Your weights stay home. The training data, the checkpoints, and the inference all live on hardware you own.
llama-factory // train gemma 4
# fine-tune on your own data, on your own GPUs
$ llamafactory-cli train \
    --model_name google/gemma-4 \
    --dataset timestamp_essays \
    --finetuning_type lora \
    --output_dir ./gemma4-in-my-voice
✓ checkpoints written locally
# serve it back to your own services
$ vllm serve ./gemma4-in-my-voice
✓ OpenAI-compatible endpoint, on your box
Period 4 — After School

Work While You Sleep

Cron turns the PDC from a tool you poke into a system that runs on its own. Schedule your hosted models to do the work, write results back to your services, and hand out a key to anyone you choose — without giving away the box.

OP 04

// AFTER HOURS — SET IT AND WALK AWAY

Rig It to Run Itself

Cron turns the PDC from a tool you poke into a system that runs on its own. Schedule your hosted models to do the work, write results back to your services, and hand out a key to anyone you choose — without giving away the box.

  • Cron + your own models. Schedule jobs that call the models you host — research, summaries, monitoring — no per-token meter running.
  • Talk to your services. A job reads from and writes to the services on your box; results are waiting when you wake up.
  • Reachable when you want. Services on the open web (on your terms) mean the jobs reach them and so do you, from anywhere.
  • Share selectively. Give one service to a family member, a colleague, or an LP — a single key, not the whole rig.
crontab // the overnight shift
# 6am: local model writes the morning report,
# posts it to a service on your own box
0 6 * * *  curl localhost:8000/v1/chat \
            -d @prompt.json \
            | curl -X POST https://pdc.local/report -d @-
✓ runs every morning, on your hardware
# share one service, keep the rest private
$ start-cli access grant report --to lp@example.com
scheduledself-hosted inferenceone key, one serviceno per-token meter
Extra Credit

Tools of the Trade

Every rig comes down to five things you procure: a box for StartOS, GPUs for your models, a VPS to front the tunnel, keys to the frontier models, and the open weights themselves. Here's my kit — and the field substitutes.

APPX A

// REQUISITIONS — WHAT'S IN THE BAG

Tools of the Trade

Every rig comes down to five things you procure: a box to run StartOS, GPUs to host your models, a VPS to front the tunnel, keys to the frontier models, and the open weights themselves. Here's my kit — and the field substitutes if you're working with what's on hand.

KIT 01 · THE BOX

Hardware StartOS runs on — your always-on server.

MY KITStart9 Server One — StartOS preflashed, plug in and go. US: Solo Satoshi.

Field substitutes

  • Start9 Server Pure (fully open firmware) or Server Lite
  • Any x86-64 mini PC — Intel NUC, Beelink, Minisforum
  • Refurb business desktop — Dell OptiPlex, Lenovo ThinkCentre
  • A VM via Proxmox — StartOS ships x86-64, ARM64 & RISC-V images

KIT 02 · THE GPUS

The silicon that fine-tunes and serves your own models.

MY KITDual NVIDIA DGX Spark, linked over ConnectX-7 — ~256GB combined memory, models up to ~405B.

Field substitutes

KIT 03 · THE VPS

The public node Start-Tunnel fronts, so your home IP stays hidden.

MY KITDigitalOcean — a cheap droplet is all the tunnel needs.

Field substitutes

KIT 04 · THE AI KEYS

Frontier models for first builds, audits, evals, and patches.

MY KITAnthropic Claude API (Opus) — the heavy lifting and the audit.

Field substitutes

KIT 05 · THE MODELS

Open weights you download, host, and fine-tune yourself — nothing leaves the box.

MY KITGemma 4 plus several others — pulled from Hugging Face and run on the Sparks.

Sources & substitutes