The Builder's Kit.
The professional tier: not a simplified version of anything, but the actual method a one-person AI-leveraged development practice runs on. Claude Code in a terminal, the written scaffolding that turns AI from a guesser into a colleague, architecture judgement, real hosting and DNS, security that fits your actual threat model, and the verification habits that make it all trustworthy. Nine documents, nothing held back. And the kit sets itself up: hand the zip to Claude Code and it interviews you, writes your project's scaffolding, builds the first increment, then hands you the controls.
Annual because the ground moves: the fee buys the current edition plus every re-issue for a year. Your downloads are yours forever; lapse and you keep every file, you just stop receiving updates. 30-day refund, and the year's fee comes off the invoice if you hire me within it.
Inside
Nine documents, in order.
1. Start here & 2. The method
The licence in plain words, then the doctrine everything implements: the spec
is the work, AI is a crew not an oracle, verify adversarially every time, judge
the result not the code, and the working loop that assembles them. Eleven years
of analysis discipline, compressed.
3. The workbench
Install Claude Code and understand what it is, the ten terminal moves that
cover the whole kit, and git underneath everything so no mistake is ever
permanent. One honest afternoon, including the subscription costs said out
loud.
4. The scaffolding
The kit's centre of gravity: four specially written .md files with full
templates. CLAUDE.md standing orders, SPEC.md in checkable statements, DESIGN.md
so the site gets your taste instead of AI's default, and the decision log that
ends re-litigation. Prompts are what you say; scaffolding is what stays
said.
5. Architecture
The boring-stack doctrine, a four-rung ladder from static site to framework
application with what forces each step, WordPress positioned honestly, a
decision table, and the structural habits (derive don't duplicate, fail loudly)
that keep a system maintainable by one person.
6. The build
Component by component: skeleton, layout and navigation, pages and content,
the form and its serverless function specified like an adult, search visibility
and schema, performance, and the pre-ship adversarial review. Each with the
instruction pattern and the verification step before the commit.
7. Hosting & infrastructure
DNS records actually explained (A, CNAME, MX, TXT, nameservers), deploy-from-
git platforms, preview deployments, environment variables and secret rotation,
email that actually arrives (SPF, DKIM, DMARC), and the infrastructure register
with a quarterly drift audit.
8. Security
Your actual threat model (weather, not snipers), protections ranked by
real-world frequency starting with accounts, form abuse handled gracefully,
the email fraud that actually costs Australian small businesses, backups with
a recovery drill, and explicit permission to ignore the enterprise theatre.
9. Ship & keep
Launch verification, the weekly/monthly/quarterly operating rhythm, growing
the system without wrecking it, and the honest edges where the method thins out
and hiring becomes the method working, not failing.
Plus: the kit installs itself
The zip carries a
setup/ folder written for the AI rather than
for you: once Claude Code is installed, one sentence ("Read
builders-kit/setup/START-HERE-CLAUDE.md and follow it") has it interview you
one question at a time, write your CLAUDE.md, SPEC.md, DESIGN.md and decision
log from your answers, set up git, and build the first increment with you
watching. Then it deliberately stops and teaches you the loop, because the
promise is capability, not a done-for-you site. Read the kit, or run it;
both are first-class.Fair warning
Who it's for, and not.
For you if
You are a founder, operator, analyst or marketer going technical, or a
developer who wants the AI-era working method; you are comfortable in a
terminal or willing to spend an afternoon becoming so; and you want capability
in-house rather than a one-off site.
Not for you if
You want a website with the least possible learning: that is the
Website Kit, it costs
less, and it is the honest recommendation. Or your project involves regulated
data, direct payment handling or someone else's ruined codebase, which are
past any kit's rung by design.
Why trust the method
Because you are reading its output. This site, its 80-odd pages, its tools
and the compilable reference build were produced with exactly the method the
kit documents, and
how I work with AI is public.
The kit does not describe a theory; it describes a practice you can
inspect.
Fair questions
Before you buy.
Why annual, when the Website Kit is one-off?
Because this kit's subject moves under it: AI tooling, hosting products and
security guidance all drift within a year, and keeping it current is real
recurring work. The fee funds the year's re-issues. The Website Kit teaches
decisions and a directing method, which age slowly; different content, different
honest price shape.
What happens if I don't renew?
Nothing dramatic, on principle: every file you downloaded stays yours and
keeps working. There are no licence keys, no expiring links, no phone-home
checks in the documents. Renewal buys the next year's editions, not continued
permission to use what you already have.
Is this a Claude Code course?
It is a working-method kit that standardises on Claude Code as its terminal
agent, installs it properly, and builds the scaffolding conventions around it.
The method (spec first, crew not oracle, adversarial verification) transfers to
any capable agent; the judgement documents are tool-agnostic on purpose.
What does "the kit sets itself up" actually mean?
The zip includes setup instructions addressed to Claude Code itself, plus
the scaffolding templates as ready files. You install Claude Code (document 3),
unzip the kit into your project folder, and tell it to read the setup file. It
then runs the kit's own onboarding: interview, scaffolding files, git, first
increment, each phase stopping for your approval, no facts invented. It ends by
handing you the controls and two example next instructions for your actual
project.
Terminal? I've never opened one.
Document 3 assumes exactly that and gets you genuinely working in an
afternoon: ten moves, git as a two-word habit, and the AI itself as your standing
helper for every error message. Willingness is the requirement, not
experience.
Will this replace hiring a developer?
For a real class of systems, honestly, yes; that is the tier's promise and
the architecture ladder tells you which class. The kit is equally honest about
its edges: regulated data, rescues, and hours that do not exist. If you hit an
edge and hire me within your licence year, the fee comes off the invoice.