Bitcoin
I save in Bitcoin. This is where I explain why and collect the resources I would share with someone curious. — Why I save in Bitcoin
What I’m using, learning, and spending time on. AI tools, Bitcoin, my church, and the projects I’m building. I update this page when something changes.
Updated
Faith, Bitcoin, and the things I keep coming back to.
I save in Bitcoin. This is where I explain why and collect the resources I would share with someone curious. — Why I save in Bitcoin
My church. Livestreams and teaching are available on YouTube. — Watch Truth Chapel
Opus 5.5 is my default, and Fable 5.2 is in daily use with it. Muse Wonder and Grok 4.7 are occasional choices, with Grok mainly coming through Grok Bot or Cursor. The one local model, Qwen3.8-27B, runs at Q4_K_M quantization through llama.cpp.
anthropic · model · default
My default. The model behind most of my Claude Code sessions — planning, larger edits, and the long, context-heavy, repo-shaped work.
+ Strong at planning and long sessions
+ Tight, deterministic output on routine edits
anthropic · model
Also in daily use, alongside Opus 5.5.
meta · model · opencode
Meta’s model, run through OpenCode. It is not a frontier coding model, but it is better than that sounds — I like how it has worked and how it responds, so it stays in the rotation for occasional use.
+ Better than its tier suggests on real work
+ A response style I keep coming back to
xai · model
An occasional third perspective, mainly when Grok Bot or Cursor uses it. Opus 5.5 is the default, with Fable 5.2 also in daily use.
+ Another perspective outside the Claude models
qwen · model · llama.cpp · local
The only local model I still run. Qwen3.8-27B at Q4_K_M quantization, served by llama.cpp on the 3090.
+ Private and local
The agent loop itself — the thing that takes a model, gives it tools, and lets it inspect a repo, edit, and run checks. Claude Code carries the daily work; OpenCode covers other providers and my local model. Codex is an occasional second read. I drive them from T3 Code.
anthropic · cli · coding agent
My default harness for high-context, repo-shaped work — planning, larger edits, and longer sessions where keeping the whole repo in view matters. Runs the Claude models; I drive it from T3 Code.
+ Strong at planning and repo-shaped work
+ Holds context across longer sessions
open source · cli · coding agent
The open-source harness, and the one I run Muse Wonder through. It is also how the local model fits the loop: OpenCode talks to llama.cpp on the RTX 3090 PC, which serves Qwen3.8-27B at Q4_K_M quantization, so the same inspect, edit, and check workflow runs against a model on my own hardware.
+ Vendor-neutral — cloud and local models in the same loop
+ Runs the 3090’s Qwen3.8-27B through llama.cpp
- One more harness to keep configured next to Claude Code
openai · cli · coding agent
An occasional implementation-and-inspection pass. I drive it through T3 Code when I want a second read against Claude Code.
+ Quick into implementation and repo inspection
+ Clean inside T3 Code
T3 Code is my main coding interface. I use Grok Bot daily, and I am experimenting with Claude Duet. Muse is another experiment in that same category.
interface · multi-harness · daily driver
My main interface and where I spend most of my time. Not a harness — it sits on top of Claude Code, OpenCode, and Codex and lets me run them from one window, with my models pinned to keyboard shortcuts so switching is a keystroke instead of a context rebuild.
+ One interface over my coding harnesses
+ Switch models and harnesses without leaving the flow
spacexsi · personal agent · app
The assistant I use every day. It runs on its own cloud computer, signs into apps the way I would, and keeps working with my laptop closed. Occasional Cursor work happens through here rather than in a separate harness I sit in. It comes with my X Subscription.
+ Keeps working without my machine on
+ Dispatches Cursor cloud agents
anthropic · personal agent · app
An assistant I am experimenting with. An ongoing thread with Claude that can work in connected apps, keep going while I am away, and come back when it needs my attention or approval. — Claude Duet
+ Carries over my Claude memory, skills, and connected apps
+ Keeps working on its own cloud computer
meta · personal agent
Meta’s personal agent. Another interface I am experimenting with alongside Grok Bot and Claude Duet.
+ Another personal assistant to compare against Grok Bot and Duet
- Too new to have an opinion yet
The paid plans that make the stack work — a different kind of thing from the agents and models they unlock.
spacexsi · Super tier · $100/mo
SpaceXSI’s bundle of Grok, Cursor, and X, on the Super tier at $100/mo. I am a heavy Grok Bot and Grok user, with occasional Cursor use on the same plan.
+ One plan covers Grok, Grok Bot, Cursor, and X
+ Replaces what used to be three separate subscriptions
anthropic · $100/mo
My $100/month Claude plan, now simply called Max. It supports my Claude Code work through T3 Code and my experiments with Claude Duet.
+ Room for longer Claude Code runs
+ $100/month in API credits for my own builds and API-backed third-party tools, including dictation software
openai · $100/mo
My $100/month plan for occasional Codex use.
+ $100/month in API credits for my own builds and API-backed third-party tools, including dictation software
Where the work and the sites actually live — the Mac, the home GPU PC, the always-on VPS, hosting, and DNS.
apple silicon · local environment · hardware
Where the work actually happens. Not an AI tool, but the machine the harnesses edit on and the checks run against, and the default over the VPS for nearly everything. Still plenty of headroom for the whole stack of CLIs, editors, and a browser at once.
+ Handles the full local toolchain comfortably
+ Quiet, cool, and always ready
- Local model inference runs on the 3090 PC instead
hostinger · KVM 4 · Ubuntu 24.04
My always-on Linux box — a Hostinger KVM 4 on Ubuntu 24.04. Workers handle all of my public web apps, so this server has one job: an always-on T3 Code host I can reach from anywhere through T3 Connect or Grok Bot, plus a Tailscale exit node. No public web ports are open; SSH only listens on Tailscale. — my referral
+ Always-on T3 Code host when my Mac is closed
+ SSH over Tailscale only — no public ports open
+ Doubles as a Tailscale exit node with automated daily backups
- I own the patching, backups, and uptime myself
Workers · R2 · Access · DNS · domains
One platform underneath my sites: Workers hosts the public apps, including this one and morningportion.com; R2 holds files and object storage; Access keeps private apps behind a login; DNS and Registrar handle my domains. Next.js apps deploy through OpenNext, while simpler sites ship as plain assets.
+ Hosting, storage, private-app access, and domains in one dashboard
+ OpenNext for Next.js; plain assets for simpler sites
- The Workers runtime is not Node, so some npm packages need a compatibility flag or a swap
nvidia · 24 GB VRAM · llama.cpp · hardware
My Windows desktop with a 24 GB RTX 3090. It runs llama.cpp serving Qwen3.8-27B at Q4_K_M quantization over Tailscale so the Mac can use the local model without giving up laptop memory.
+ 24 GB of VRAM handles 30B open-weight models comfortably
+ Shared with the Mac over Tailscale
- Only on when I need local inference or gaming
Companion utilities and helper surfaces that support my workflow.
voice · dictation · macOS · Windows · Linux · open source
Free, open-source local dictation from Anomaly, the team behind OpenCode. It replaced Superwhisper for me: hold a shortcut, speak, and it transcribes on the device with Parakeet or Cohere models and pastes into the app I am using. Post-processing is optional, and it can run through the Claude Code and OpenAI subscriptions I already pay for. — Hex
+ Runs on Mac, Windows and Linux at no cost
+ On-device transcription with Parakeet and Cohere models
+ Post-processing through my existing Claude Code and OpenAI subscriptions when I want it
vpn · network
Zero-config mesh VPN that keeps my machines and VPS reachable securely.
+ Simple secure networking across devices
+ Works well for remote agent access