By the editorial team · Published October 26, 2026 · 19 min read

Example 01 / Publish

Blog post: from a bookkeeper's notes to a published post, annotated

A vibe writing example, annotated: a bookkeeper's raw notes, her prompts, the AI draft with its tells, and the 620-word post she published. Every edit shown.

Read the layers4 layers · 22 margin notes · Blog post, about 700 words

Writer: A self-employed bookkeeper, nine years in, writing under her own name on her own site
Format: Blog post, about 700 words
Tools used: A chat assistant
Session length: About 50 minutes in the chat, then 35 minutes of revision by hand the next morning
Prompts used: Blog Post mega prompt (the session brief), Bounded Edit (one turn), the 'final check' turn from the Vibe Writing Mega Prompt
Turns: 5: brief, answers and draft request, bounded edit, cut, final check

This is the whole trail of one short blog post: the notes a self-employed bookkeeper typed into her phone over three weeks, the brief she gave a chat assistant, the first draft that came back, and the 620 words she published after an evening and a morning of revision. Nothing has been tidied. The draft still has its [VERIFY] marker and its numbered list.

She is a good writer who does not write often. Her site has about a dozen posts, all first person, all about running a one-woman practice. She used the assistant because she had the numbers and the point and had been failing to start for three weeks. What the assistant gave her was a start she could argue with.

Read the layers in order. The draft is the least interesting layer. The margin is where the work happened: what she kept, what she rejected, and the two places the draft said something she did not believe.

Layer 1You

Source notes, as she jotted them over three weeks

Blog post idea. The free consult thing. Been meaning to write this since June.

- stopped offering the free 45 min discovery call end of March. replaced w/ paid "books review", $150, one hour, I actually open their file (or their shoebox) before the call. fee comes off first month if they sign.

- why: Thursday afternoons were ALL free calls. Jan-Mar I did 27 of them. 7 became clients. so 20 calls = 15 hours + prep = people asking me how to categorize an owner draw in QB and then not hiring me. one guy took notes the whole time. fine. but I'm not a helpline.

- numbers (from my inquiry sheet, not memory): Jan 12 inquiries, Feb 9, Mar 12. = 33. 27 took the call. 7 signed.

- after: Apr 7, May 5, Jun 8 = 20. 9 paid for the review. 6 signed. 3 paid and didn't sign (one said she'd be back in the fall, we'll see).

- so inquiries down ~40%. clients down 1. hours on non-clients: way down. kept $450 from the 3 that didn't sign.

- the thing I can't figure out: Apr-June is ALWAYS slower for me. post tax season. people panic in Jan and again in Sept. so is the drop the fee or the calendar?? I don't have a clean year to compare bc 2025 I wasn't tracking properly. honestly don't know. don't want to pretend I do.

- client remark (the dog groomer, don't name her): something like she almost didn't book bc of the fee, then figured anyone who charges for the first hour probably doesn't waste it. that stuck w me.

- what I'm NOT saying: that everyone should charge. my friend who does payroll for restaurants says free calls are her whole pipeline. different clients.

- also the dumb part: I was scared for like 2 weeks in April when the phone went quiet. nearly put the free call back.

- tone: like my other posts. I say "I" a lot. no "as a business owner". no listicle.

- end on the doubt probably? or on the groomer? not sure

Keptfrom my inquiry sheet, not memory
This parenthetical became the fact rule for the whole session and, later, a sentence in the post. Every number the assistant used had to trace back to that sheet.
CheckedJan 12 inquiries, Feb 9, Mar 12. = 33.
She reopened the spreadsheet before pasting the notes. February had been 10 in her memory and was 9 in the sheet. The sheet won.
Kepthonestly don't know. don't want to pretend I do.
The doubt she never resolved. She flagged it in the brief as the one thing the draft must not fix, and it is the thing the draft fixed anyway.
Changedthe dog groomer, don't name her
In the brief she went one step further and dropped 'dog' as well. Her town is small, and two of her other clients know the groomer. The post says 'a groomer'.
KeptI'm not a helpline.
The one line she knew would survive in some form. It survived verbatim, which is more than the assistant's version of it managed.

Layer 2You

The direction she gave, turn by turn

Turn 1. Pasted the Blog Post mega prompt with the blocks filled in. The parts that mattered:

Writer: a self-employed bookkeeper, nine years in, mostly one- and two-person service businesses. Writing under my own name on my own site.

Writing: a first-person post, about 700 words, on why I stopped offering a free first consultation and what happened to my inquiries over the next three months.

Reader: other solo bookkeepers and accountants who have thought about charging for the first call and are scared the phone will go quiet. They read on their phones, at night.

It must: give them my real numbers and my real doubt and let them decide. It must not tell them to do it.

Raw material: [the notes above, pasted whole, plus the monthly totals from the sheet].

Voice: I say "I" a lot. Short paragraphs. I admit when I don't know something. I never open with a question.

Banned: no listicle, no "as a business owner", no "journey", no "value your time", no closing summary, no bullet lists in the body, no sentence that starts with "Here's the thing".

Definition of done: every number matches my sheet. The doubt about the calendar stays a doubt; do not resolve it. It sounds like my last three posts. Ask me up to three questions first.

Turn 2. Answers: 1. Credited against the first month, not refunded. If they don't sign, I keep it. 2. The inquiry sheet is a spreadsheet, one row per inquiry, with where they came from and what happened. 3. Use the groomer remark as a paraphrase. No name, no "dog", just "a groomer". Draft it. One draft, nothing else.

Turn 3. You resolved the doubt. "The numbers speak for themselves" and "the fee filtered out" are the opposite of what I told you. Change only the paragraph that starts "The numbers speak for themselves." Put my doubt back as a doubt: it might be the fee, it might be the calendar, I can't tell because I wasn't tracking last year. OLD and NEW, then wait.

Turn 4. Cut everything from "What I'd tell anyone considering it" to the end. I don't want a lesson. I'll end it myself.

Turn 5. Final check.

AddedAsk me up to three questions first.
The questions caught one thing she would have gotten wrong on the page: whether the $150 is refunded or credited. Her notes said 'comes off first month' and nothing about the other case.
Keptno "as a business owner"
Carried from her notes into the banned list word for word. The assistant obeyed it. Banned lists work for phrases, but they work less well for moves, which is the lesson of turn 3.
AddedThe doubt about the calendar stays a doubt; do not resolve it.
A stated constraint, in the definition of done, and the draft broke it anyway. This is why the judge move exists. A rule in the brief raises the odds. It does not guarantee anything.
ChangedChange only the paragraph that starts "The numbers speak for themselves."
She chose a bounded edit for the meaning error instead of regenerating the whole draft. The OLD/NEW came back clean, and she still rewrote that paragraph by hand the next morning, because the draft's framing had leaked into three other paragraphs.
RejectedI don't want a lesson. I'll end it myself.
She cut the summary rather than asking for a better one. Asking for a better summary would have gotten her a better summary, which was not the problem.

Layer 3AI proposes

What the assistant returned, including the first draft verbatim

Turn 1 reply. I understand this as a first-person post of about 700 words in which you explain why you replaced a free 45-minute discovery call with a paid books review, and report what happened to inquiries and sign-ups over the following three months, written for solo bookkeepers who are considering the same change and are nervous about it.

The raw material is missing what happens to the $150 if the person does not sign. It says the fee "comes off first month" but not whether it is otherwise refunded.

1. Is the $150 credited against the first month only, or refunded if they do not become a client? 2. What is the "inquiry sheet"? Readers may trust the numbers more if I can say what it is in one phrase. 3. The groomer remark: may I use it as a paraphrase, and how much detail about her business is allowed?

Turn 2 reply (the draft):

Why I Stopped Offering Free Consultations (And What Happened Next)

Let's be honest: the free consultation is practically a rite of passage for anyone who runs a service business. For most of my nine years as a bookkeeper, I offered one too. A free 45-minute discovery call, no strings attached. It felt generous. It felt like the right way to start a relationship.

At the end of March, I stopped. In its place, I introduced a paid discovery session: one hour, $150, and I review their books before we meet. If they become a client, the fee is applied toward their first month. If they don't, the hour was still worth something to both of us.

Here's what I learned over the next three months.

The problem with free

For a long time, my Thursday afternoons were devoted entirely to free calls. Between January and March, I took 27 of them. Seven of those conversations turned into clients. The other 20 were, in effect, free advice sessions. People wanted to know how to categorize transactions in QuickBooks or whether a certain expense was deductible, and then they went off and handled it themselves.

I don't blame them. But I eventually realized that my expertise had real value, and that giving it away was not sustainable.

The numbers

I track every inquiry in a spreadsheet, so I can share real figures rather than impressions.

Before the change, from January through March, I received 33 inquiries. Of those, 27 booked a call and 7 signed on. That's a conversion rate of roughly 21 percent.

After the change, from April through June, I received 20 inquiries. Nine of them paid for the review, and 6 became clients. That's a 30 percent conversion rate on inquiries, and a 67 percent conversion rate on paid sessions.

In other words, inquiries fell by about 40 percent, but I lost only one client compared to the previous quarter. I also spent dramatically fewer hours with people who were never going to hire me, and I earned $450 from the three sessions that didn't convert. Industry data suggests that paid consultations typically convert at significantly higher rates than free ones [VERIFY: source and figure for consultation conversion benchmarks], and my experience lined up with that.

The numbers speak for themselves. The fee filtered out the people who were only looking for free help and left me with serious prospects who were ready to commit.

What clients said

One of my new clients told me something that stuck with me. She said she almost didn't book because of the fee, but then realized that anyone who charges for the first hour probably doesn't waste it. That's exactly the kind of client I want to work with: someone who values expertise and comes prepared.

There were a few quiet weeks in April when the phone didn't ring much, and I'll admit that was uncomfortable. But I trusted the process, and by June the inquiries were picking back up.

Is this right for everyone?

Probably not. A friend of mine handles payroll for restaurants, and free calls are the backbone of her pipeline. Her clients have different expectations and different budgets. The right approach depends on your niche, your pricing, and the kind of relationship you want with clients.

What I'd tell anyone considering it

If you're on the fence, here are three things I'd keep in mind:

1. Track your inquiries before you change anything, so you'll know what actually happened. 2. Make the paid session genuinely useful. Open their books before the call. 3. Expect a quiet stretch at first, and don't panic.

At the end of the day, charging for the first hour gave me more than time back. It changed the kind of conversations I was having. Three months in, I have fewer inquiries, almost the same number of clients, and my Thursday afternoons back. I wouldn't go back.

Turn 3 reply (bounded edit). OLD: The numbers speak for themselves. The fee filtered out the people who were only looking for free help and left me with serious prospects who were ready to commit. NEW: I want to be careful with these numbers. April through June is always my slow stretch, so the drop in inquiries might be the fee, or it might be the calendar, or some of each. I was not tracking inquiries properly last year, so I have no clean comparison. I genuinely do not know.

Turn 5 reply (final check). Claims that would need a source: the sentence beginning "Industry data suggests", marked [VERIFY], has no source in your material. Places I may have changed your meaning: "my expertise had real value" (your notes say Thursday afternoons and "not a helpline"); "by June the inquiries were picking back up" (June was 8 after May's 5; "picking back up" is my reading). Sentences that may still sound generated: "Let's be honest", "Here's what I learned", "no strings attached", "In other words".

RejectedThe numbers speak for themselves. The fee filtered out the people who were only looking for free help
The draft resolved the doubt she had told it not to resolve, and resolved it in the direction that flatters the decision. This is the quiet meaning change: her 'I don't know' became 'the numbers prove it'.
RejectedBut I trusted the process, and by June the inquiries were picking back up.
She nearly put the free call back. 'Trusted the process' is the opposite of what happened, and the final check did not catch it. She caught it reading aloud.
Changedmy expertise had real value, and that giving it away was not sustainable
Her reason was Thursday afternoons and a help desk, not self-worth. The draft turned a scheduling complaint into a statement about value. She wrote the negative version of this sentence into the post to get it out of her head.
Checked[VERIFY: source and figure for consultation conversion benchmarks]
She had no such source and did not go looking for one. She deleted the whole sentence. The marker did its job by making the sentence easy to find and easy to cut.
CheckedThat's a conversion rate of roughly 21 percent.
All three percentages traced correctly to her sheet (7 of 33, 6 of 20, 6 of 9). She kept the counts and cut the percentages anyway, because her readers do that division in their heads and the counts are more honest about the sample size.
KeptBefore the change, from January through March
The draft's decision to give the before and after figures as two matched blocks, in the same order, survived into the final. That shape was the assistant's and it was better than her notes.

Layer 4You

What she published, after revision by hand

Three months without the free call

At the end of March I stopped offering a free 45-minute discovery call. I replaced it with a paid books review: one hour, $150, and I open their file before we talk. If they sign, the $150 comes off their first month. If they don't, I keep it.

I did not do this because I decided my time was worth more. I did it because Thursday afternoons had turned into a help desk. From January through March I did 27 of those free calls. Seven of the people on them became clients. The other 20 wanted to know how to categorize an owner draw in QuickBooks, or whether the truck was deductible, and then they went off and did it themselves. One of them took notes the whole time. Fair enough. But I am not a helpline, and I had started to dread Thursdays.

Here are the numbers, from my inquiry sheet and not from memory. January, 12 inquiries. February, 9. March, 12. Thirty-three in total. Twenty-seven took the call. Seven signed.

April, 7. May, 5. June, 8. Twenty inquiries. Nine paid for the review. Six signed. Three paid, got their hour, and did not sign, though one said she would be back in the fall and I half believe her.

So: inquiries down about 40 percent. New clients down by one. Hours spent with people who were never going to hire me, down from something like 15 to about 3. And $450 I would not otherwise have, from the three reviews that went nowhere.

That looks like a clean win. I want to be careful here, because I do not think it is clean.

April through June is always my slow stretch. People panic about their books in January and again in September. In May they are mostly fine. So the drop might be the fee, or it might be the calendar, or some of each. I would love to put this spring next to last spring, but last spring I was not tracking inquiries properly, so there is nothing to put it next to. I genuinely do not know, and I am not going to pretend to in a blog post.

What I can say is that the people who came through were different. One client, a groomer, told me she almost did not book because of the fee, and then figured that anyone who charges for the first hour probably is not going to waste it. That stuck with me. It was true of me, and I think it was true of her: by the time we got on the call she had already pulled her bank statements.

I should also say that the first two weeks of April were bad. The phone went quieter than it had been since my second year, and I had the old "free consultation" page open in my site editor, ready to put it back. I did not, mostly out of stubbornness. If May had stayed at five, I think I would have.

I am not telling you to do this. A friend of mine does payroll for restaurants, and free calls are her whole pipeline. A restaurant owner who has to pay to meet her will meet someone else. Her clients are not my clients. Mine are mostly one- and two-person service businesses who have been avoiding their books for a year and know it.

So this is where I have landed, which is nowhere tidy. Fewer inquiries, one fewer client, my Thursdays back, and a question I cannot answer yet. Next spring I will at least have two Aprils to put side by side. Both with the fee. That will not settle it, but it is more than I have now.

ChangedI did not do this because I decided my time was worth more.
She wrote the negative of the assistant's framing because the framing was in her head now and she wanted it out of the piece. The sentence that follows is the reason from her notes, nearly verbatim.
AddedI had the old "free consultation" page open in my site editor, ready to put it back.
Not in the notes, which only said 'nearly put the free call back'. She added this by hand the next morning. It is the most specific sentence in the post and the one readers quoted back to her.
RejectedSo this is where I have landed, which is nowhere tidy.
This replaces the draft's three-item advice list and its motivational close. Her notes had asked 'end on the doubt probably?' and she did, with the groomer moved up where she belonged, as evidence rather than as a landing.
KeptFair enough.
A two-word sentence restored from her notes ('fine.'). The draft had smoothed that beat into 'I don't blame them.' The uneven rhythm is hers and she put it back on purpose.
Checkeddown from something like 15 to about 3
She did this arithmetic herself: 20 non-client calls at 45 minutes, then 3 non-client reviews at an hour. Prep time was never tracked, so she hedged with 'something like' rather than let the figure sound measured.
ChangedNext spring I will at least have two Aprils to put side by side. Both with the fee.
The draft ended with 'I wouldn't go back.' She does not know that. This is the true version: a plan to find out, and an admission that even that will not settle it.

What the example shows

A stated constraint is not a kept constraint

She wrote 'do not resolve the doubt' into the definition of done, and the draft resolved it in the first paragraph after the numbers. The brief raised the odds, but it did not make the rule hold. The judge move exists because a draft can obey nine instructions and break the one that matters, in confident prose that reads as finished.

A [VERIFY] marker is a question, not an errand

The draft wanted a benchmark and marked it honestly. She did not go looking for one. She asked whether the post needed the claim at all, decided it did not, and cut the sentence. Most markers resolve that way. The ones that survive are the claims the piece cannot do without, and those get a real source or get rewritten as opinion.

Bounded edits fix sentences, but they do not fix frames

The OLD/NEW on the doubt paragraph came back exactly right. But the draft's frame, that the fee filtered out the wrong people, had already leaked into the client section ('exactly the kind of client I want') and the April section ('trusted the process'). She found those only by reading the whole post aloud the next morning, which is the argument for a final human edit after the last bounded edit.

Keep the assistant's structure when it is better than yours

Her notes gave the numbers in one tangled bullet. The draft laid them out as two matched blocks, before and after, in the same order. She kept that shape and rewrote every sentence inside it. Ownership is about the decisions, and keeping a good structure on purpose is a decision.

The best sentence arrived after the chat was closed

The site-editor detail, the one readers quoted back, was written at the kitchen table with the laptop shut. The session got her a draft to argue with and a shape to pour into. The sentence that made the post hers did not come from a turn.

The ledger: who did what

An honest tally. The point is not that the human did more. It is that the human did the parts that carry the authorship.

You

The idea, the decision to write it, and the three weeks of notes
Every number, checked against the inquiry sheet before and after drafting
The doubt about the calendar, and the decision to leave it unresolved in print
The groomer remark, paraphrased and anonymized a step further than the notes
The reason for the change (Thursday afternoons), restored after the draft reframed it
The site-editor detail and the ending, written by hand with the chat closed
Deleting the benchmark sentence rather than sourcing it
The read-aloud pass that caught 'trusted the process', and roughly two thirds of the final sentences

AI proposes

Three pre-draft questions, one of which caught the credited-versus-refunded gap
The first full draft, about 650 words, in one turn
The two-block before/after layout for the numbers, which she kept
The conversion percentages, which she checked and then mostly cut
One OLD/NEW bounded edit on the doubt paragraph, later rewritten by hand
A final-check list that caught the [VERIFY] sentence and two meaning changes, and missed a third
The title, the subheadings, and the three-item advice list, all rejected

Questions

How much of the final post did the assistant write?

By her own count, about a third of the final sentences are unchanged or lightly edited from the draft, nearly all of them in the two numbers paragraphs. The argument, the doubt, the reason for the change, the April section, and the ending are hers. The structure of the numbers section is the assistant's, kept on purpose.

Why did she not just ask the assistant to fix the doubt paragraph and keep the rest?

She did, with a bounded edit, and the paragraph came back right. The problem was that the draft's frame had spread into other paragraphs that were not wrong on their face. A bounded edit cannot find leaks it was not pointed at. Reading the whole post aloud did.

Is it dishonest to publish this without saying an assistant helped?

That depends on where it appears and what the readers expect. Her site has no stated policy and her readers are other bookkeepers, so she chose a one-line note on her about page rather than a disclosure on every post. The ethics lesson covers how to decide. The short version is to follow the rules of the venue and to never let a draft stand in for facts you have not checked.

Should I use the Blog Post mega prompt or the universal one for a post like this?

The Blog Post mega prompt. It carries the same fact, voice, and turn rules as the universal brief, with format-specific additions for an opening, a reader, and a length. The universal brief is the right choice when the format is unusual or when you plan to write several different things in one session.

Next exampleShort story: from a fragment to 800 words of flash fiction, annotatedFlash fiction, about 900 words

Keep going