Built with Muse
Built with Muse · In public, from day one
Built with Muse.
Real builds with Muse, in public, from day one. What worked, what broke, and what it can do for your business.
This page is not a demo, and it is not theory. It is the record of one builder wiring Muse, Claude Code, and Codex into a real working system, and learning how to use them while everyone else is still learning too. It is updated as we build, not once and left alone.
See the latest builds ↓ Work with Genius Loci →
Why listen to this page
Because it is specific, honest, and early. Most AI content right now is either a polished demo, or a warning with no alternative. This page is neither. It shows a real stack, real mistakes, and real fixes, with dates, so you can check the work and copy the parts that fit your business.
If you want to learn how to use Muse, this is the place to watch someone do it, in public, from the start.
The stack, in plain words
Muse — the top layer
The single point of contact. Dominic talks to Muse. Muse brings plans, drafts, and verification. Dominic makes every call.
Claude Code — the builder
Long, careful repo work. Frontier quality, session-capped, used where it matters most.
Codex — the second builder
Small, tightly scoped work, and cross-review. A second set of eyes, not a second system.
Local model — the free worker
A model on the home desktop, free and unlimited, inside limits. It does the bulk work, with review before anything lands.
Rule for every build: a ticket with objective, context, constraints, and acceptance criteria. A diff reviewed before it lands. A scheduler registration, so nothing dies after one manual run. A $1/day cloud ceiling, on purpose.
The elephant in the room
Yes, AI can be dangerous. A lot of the public pushback is not about fear of change. It is about real risk, and real misuse, and that concern is fair.
That is why this system is built with guardrails, in public: nothing publishes on its own. Nothing spends on its own. No secrets are in reach of the agents. Every build is reviewed before it lands. Every social post goes through approval. A person stays in charge, and the work is visible.
The argument here is not “trust AI.” It is “here is how you use it with guardrails, so you can check the work.”
The lanes
New content on this page falls into four lanes. When a lane gets a new piece, it goes in the build log below, in date order, newest first.
1. The integration story
How Muse, Claude, and Codex were wired together, step by step, and what each one is actually for.
2. Proof of work
A real build, fixed, shipped, or verified, with before and after. No invented clients, revenue, or results.
3. How to use Muse
One usable lesson per piece, that a reader can copy. A prompt, a queue, a check, a rule.
4. Guardrails
AI risk, pushback, and the controls that keep a person in charge. Addressed directly, not as a disclaimer at the end.
Build log — updated as we build
Newest first. This log is the page's heartbeat. If it stops, the page is stale, and that is a signal, not a detail.
2026-10-25 · Proof of work
The homepage, reconciled, not rebuilt
The first homepage fix stripped the theme and its hover animations. The reconciled version kept every section and animation, shrank the headlines to a readable size, put AI and Marketing at the front of the menu in a logical order, removed repeated products, took the Object in Focus card away, and put a real intake form at the bottom. Lesson: preserve what works, fix what is broken, verify it live.
2026-10-25 · How to use Muse
From idea cards to a nightly work queue
Idea cards sat there with no next step. Now, every night at about 2:23 AM, Muse turns unpublished drafts, acted-on ideas, and unfeatured products into up to three concrete work items, each with an objective, a format, a hook, a CTA, and what it needs. Drafts only, for morning approval. Nothing publishes on its own.
2026-10-25 · Proof of work
From Instagram DM to a real intake form
Serious buyers were sent to Instagram DM. That worked for followers, not for $5,000–$20,000/month retainer clients. The services page now has a short intake form as the main route, with DM secondary, and it lets people qualify themselves by budget, revenue, offer, and decision maker. Unqualified replies still come in, tagged as not a fit.
2026-10-25 · How to use Muse
A weekly revenue leak scan, on a loop
Every Sunday at about 9:23 AM, Muse scans for revenue leaks — draft products that should be live, archived products still linked, empty collections, duplicate pages, broken links, and unanswered intake replies — and reports them as fixes to approve. Report only. Nothing fixed, published, or spent without a yes.
2026-10-25 · Integration story
This page itself
Built with Muse, as a draft, in the same hub style as the other Genius Loci pages, with a build log that grows as we build. The Facebook page, Built with Muse, will link here, not to the main homepage, so the narrative has one home.
Work with Genius Loci
If this way of working looks useful for your business, start with the intake form. It is short, on purpose, and it is how serious inquiries are qualified.
Or start with the story of the whole system: The Loci System, The Index, and Start.
Explore Genius Loci
Market Research · Founder Bio · The Index · Start · Agent Hub · The Build · Managed AI Services
Last updated: 2026-10-25. This page is updated as we build.