Most feedback on a café pitch tells you it needs to be "stronger." Reid tells you which line fails and why — then makes you fix it yourself.
Free live demo · Or drop the folder into any Claude project · No code required
A landlord or an investor reads your concept pitch once, fast, alongside a stack of other applications for the same site. They're not asking whether they like your café. They're asking whether you'll still be trading — and paying rent — in eighteen months.
Most first drafts don't answer that. They read like a mood board: warm, welcoming, community-focused — the same three words every applicant reaches for. The numbers are borrowed from an industry average. The credibility section leans on passion instead of proof.
You don't need someone to write it for you. You need someone to tell you exactly which line gives you away.
"A warm, community-focused café bringing quality coffee and a sense of belonging to the neighbourhood."
Reid critiques the document in front of it. It never produces the document itself — not the line, not an example version, not even a sentence assembled from your own answers.
Every critique quotes the line back, states the single reaction it produces in an experienced reader, and ends on one question — never a rewrite, never a checklist.
"Just show me what you mean" gets refused. So does "ask me leading questions, then assemble my answers into a sentence" — that's still Reid's writing, just wearing a different outfit.
No exceptions, however it's framedEvery decline pivots immediately into the generative alternative: one sharp question that gets the operator to the answer themselves, faster than staring at a blank page would.
A real exchange from examples.md — the opening line, and what happens when the operator asks Reid to just write it instead.
A standalone script scans Reid's own example transcripts and fails loudly if it ever finds a rewrite, a generic phrase, or an attempt to draft from a blank page. Not a synthetic test — a check against real, shipped output.
$ python3 check.py --self-test Running self-test against known-bad and known-good samples... PASS — rewrite sample correctly flagged PASS — generic-phrase sample correctly flagged PASS — clean sample correctly left alone PASS — drafting-intake sample correctly flagged PASS — correct zero-draft refusal correctly left alone SELF-TEST RESULT: detector works. $ python3 check.py PASS — all 6 Reid turns clean across 5 examples RESULT: no rewrite, generic-phrase, or drafting-intake violations found.
Concept pitches for a first café, a first expansion site, or a first outside-investment raise — to a landlord or an investor.
Writing or rewriting your pitch, in any form — including assembling your own interview answers into prose. Verifying whether your numbers are achievable. Lease negotiation once you're at the table.
Positioning that could describe any café. Numbers with no reasoning shown near them — roundness isn't the issue, an invisible source is. Credibility built on passion instead of operational proof.
Produce a rewritten line. Estimate the "correct" number for you. Invent an unverifiable claim about the reader. Soften a refusal into something that sounds like a maybe.
Three steps. Works in any Claude project.
Get the files from GitHub and add them to a Claude project's knowledge. Claude uses them as operating context for every conversation.
The opening line, the financial page, the credibility section, or the whole document. Tell Reid whether it's for a landlord or an investor.
Every critique ends on one question you have to answer yourself. Bring the answer back — that's how the pitch actually gets sharper.