The Starter Spec turns your brand voice into a machine-readable spec: eight channels scored zero to ten, plus the hard bans and behavioral rules that hold the sound in place. The finished spec is a document. You paste it into your tools once, and every draft after that starts from your voice. You build it in a 25-minute self-serve session. I engineer it, sign it, and ship it the next business day.
SELF-SERVE · 25 MINUTES · SHIPS NEXT BUSINESS DAY · SIGNED BY A HOO-MAN (ME)
REDACTED EXCERPT · WHAT A SPEC LOOKS LIKE INSIDE
The first failure is copy that sounds wrong: technically fine, recognizably nobody. The second is copy that makes things up: confident specifics your brand never lived. The Starter Spec ships both fixes together.
Your voice as a signed document: your eight-channel coordinate, the behavioral rules for each channel, your taboo words placed as hard bans, and context ranges so your promos stay punchy and your apologies stay human, from the same set of rules.
It ships in three sizes (full, compact, and micro), because ChatGPT, Claude, and Klaviyo each cap their instruction fields differently, and a spec that gets truncated is a spec that gets ignored.
What it closes: the machine finally has instructions it can follow the same way twice. The excerpt above shows the shape.
A four-tab workbook and prompt library you fill with your brand's true stories and facts: your origin, your numbers with dates, your refusals, your customer moments. Each entry is indexed by the claim it supports, so your AI can find the right story when it argues a point.
The workbook carries its own AI rules on tab one: never invent entries, never alter the load-bearing facts, and when no entry supports a claim, say so instead of improvising one.
What it closes: your AI draws on real material instead of inventing some. The spec governs how you sound. The bank supplies what is true.
THE ACTUAL INBOX MOMENT · PLAIN FILES, ON PURPOSE
The obvious question deserves a straight answer: $249 buys a file Notepad could open. That's correct, and it's engineered that way. Plain text is the one format every AI tool accepts whole, so your spec installs in ChatGPT, Claude, Klaviyo, and whatever launches next quarter. It can't expire, can't break, and doesn't phone home to my servers to stay alive. (A dashboard would just be my subscription fee wearing a costume.) The price pays for the measurement: your session read, your contradictions resolved, your rules written and signed. The document is where that engineering lives, the way a prescription is ink on a pad.
25 minutes, self-serve, on your schedule. You sort words, answer 12 calibration questions at gut speed, set the blind faders, and name the words your brand would never say.
I read your session, resolve the contradictions, and write the rules: channel behaviors, context ranges, and your taboo words placed as hard bans.
Your spec gets versioned, dated, and signed by an actual human. Me. A spec nobody stands behind is just a document.
It lands in your inbox the next business day, in three sizes, with the Context Bank template and the Install walkthroughs riding along.
One honest note on method. The Starter Spec reads one source: what you declare about your voice. It is the fastest honest reading there is, and for one person with a clear sense of their voice, it lands. The Live Session reads three sources (what you declare, what you publish, and what your gut picks under pressure) and argues out the gaps live. That difference is the whole reason both exist.
Your delivery email brings written walkthroughs for the major tools. The principle is simple: give your brand a dedicated project, paste the spec in whole, and never summarize it. From then on, your AI writes from your settings instead of the internet's average, and you stop re-prompting.
Install the spec and generate one piece with a prompt you have used before. If that first output does not sound more like your brand, reply to your delivery email within 14 days with the output and what is not working, and I revise your spec once, free. The full wording lives in the terms.
The model behind your AI updates on a random Tuesday, or six months of small edits pile up, and the copy slowly stops sounding like you. When you want certainty, you order a Spec Audit + Patch: you send fresh, unedited AI output, I score it across all eight channels against your signing-day baseline, and if the voice has moved, I patch the spec. You order it when you want it. There is no subscription and no schedule.
Both end with your voice as a signed, machine-readable spec. They differ in how many sources get read, and in how much of me you get.
| The Starter Spec | The Live Session | |
|---|---|---|
| Price | $249 founding, first 100 specs only ($500 after) | $1,997 founding ($4,000 after) |
| Format | Self-serve, 25 minutes, on your schedule | Live with me, 90 minutes |
| Readings | One: what you declare | Three: what you declare, what you publish, and what your gut picks, reconciled live |
| The machine spec | In three sizes, engineered and signed | In three sizes, engineered and signed |
| Human guide + deck | Not included | Included: a human-readable voice guide and the reveal deck for your team |
| The Context Bank | Template and prompt library, you fill it | Completed from your materials, verified, claim-tagged, signed |
| Guarantee | One spec revision within 14 days, based on your feedback | Two audit and retuning rounds within 90 days, scored on unedited AI output |
| Best for | One person with a clear sense of the voice | Brands where several people and tools write, and the drift already shows |
The Live Session has its own page, with the discovery call explained. It lives at the Session page.
A 25-minute session from you. A signed, machine-readable spec from me. The re-prompting era ends there.
THE FOUNDING PRICE RISES WHEN THE FIRST HUNDRED SPECS ARE GONE, AND RISES GET ANNOUNCED FIRST. THERE IS NO COUNTDOWN TIMER HERE, BECAUSE I DON'T DO THOSE.