We Generated a Tiny Tool Brief — Then Cut Half of It Before Building

Quick scan

• Problem: Tool ideas often start with too many features. The hard part is cutting them before building, not after.
• Approach: A single maker session using five tools in sequence — from Tarot brief to launch caption.
• What this shows: One transparent example of the Ship a Tiny Tool workflow. Illustrative, not a measured user study.
• Use this now: Run your own idea through Tiny Tool Tarot first, then bring the brief to Landing Page Lie Detector before touching a line of code.

This is a fictional but realistic example. No build metrics, no user data. The value is in seeing what the scope cuts looked like, not in the numbers.

The starting brief

The Tarot card drawn: The Parser — "A tool that reads something messy and makes it usable."

Rough brief after the card:

"A tool that takes raw meeting notes, extracts all mentioned action items, assigns owners based on name mentions, flags overdue items based on date mentions, and outputs a structured task list in Markdown."

CTA draft: "Transform your meetings."

Launch headline: "The only meeting tool you'll ever need."

What the Lie Detector flagged

The Landing Page Lie Detector pass on the draft brief and headline:

| Phrase | Flag | Why | | --- | --- | --- | | assigns owners based on name mentions | Requires verification | Name-to-person resolution needs context the tool cannot have | | flags overdue items based on date mentions | Requires verification | Date parsing needs format standardization; timezone is ambiguous | | the only meeting tool | Absolute claim | Cannot be true for all meeting contexts | | transform your meetings | Vague outcome | No mechanism; no constraint on what "transform" means |

Cuts made:

• Owner assignment removed — requires entity recognition that cannot run browser-only without errors
• Date/overdue flagging removed — ambiguous across formats and timezones
• "structured Markdown" simplified to "numbered list" — "structured" adds no information
• Headline and CTA replaced entirely

After the cut, the brief became:

"A tool that extracts a numbered list of action items from pasted meeting notes. No owner assignment, no date parsing, no blockers — those require judgment the tool does not have."

Naming pass

Cursed Title Generator output — five candidates:

1. Meeting Transcription Transformer (rejected — "transformer" overclaims the output)
2. Action Item Extractor (accurate but describes the internal mechanism, not the user action)
3. Meeting Note Reducer (chosen — describes what happens to the input)
4. Transcript Distiller (poetic but unclear without context)
5. The Blocker Hunter (rejected immediately — we removed blockers from scope)

Selected: Meeting Note Reducer

Why: A first-time visitor can parse it without a tooltip. "Reducer" is accurate — the input is long, the output is short.

CTA pass

Button Copy Casino output — six candidates evaluated against the scoped brief:

| Label | Status | Reason | | --- | --- | --- | | Transform your notes | Rejected | "Transform" still overclaims after the cut | | Extract action items | Chosen | Describes the exact action the tool performs | | Get my action list | Considered | Informal without being clearer | | Reduce meeting notes | Rejected | Describes the input change, not the user benefit | | Summarize my meeting | Rejected | Wrong framing — it produces a list, not a summary | | Clean up notes | Rejected | Does not match the scoped output |

Selected: Extract action items

Why: One verb, one object, matches the brief exactly. Does not imply anything the tool cannot deliver.

Launch caption

Screenshot Caption Gobbler output based on the final brief:

"Paste your meeting notes. Get a numbered list of what to actually do. No account needed, runs in your browser."

What was retained:

• The two-sentence structure (what you do, what you get)
• "runs in your browser" — honest constraint that is also a benefit

What was removed from an earlier draft:

• "in under 10 seconds" — no measurement was done
• "smart extraction" — added nothing
• "perfect for remote teams" — unverified audience claim

Commitment Calendar entry

Set for 2026-06-10 — two weeks after the brief was finalized.

Single check: Does the CTA "Extract action items" still match what the tool actually produces? If the output format changed, the CTA and caption need to update before traffic sees the discrepancy.

What changed, explicitly

Starting brief: 4 features, 1 overclaiming headline, 1 vague CTA.

After the maker path:

• Features removed: 2 (owner assignment, date/overdue detection)
• Features kept: 1 (action item extraction as a numbered list)
• Headline: replaced
• CTA: from 3 vague words to 3 specific words
• Caption: from 2 unsupported claims to 0

The brief did not get shorter because the tools forced cuts. It got shorter because the Lie Detector made each cut's cost visible before the build started.

What this example does not show

• Whether Meeting Note Reducer was ever built
• Whether anyone used it
• Whether the cuts improved any metric
• Whether the Tarot card "The Parser" was the right starting point

This is one fictional session designed to show what the scope cuts look like when they happen before building instead of during or after.

Related

• Run your own brief in Ship a Tiny Tool workflow
• Start the idea in Tiny Tool Tarot
• Audit the brief in Landing Page Lie Detector
• Write the CTA in Button Copy Casino
• Write the caption in Screenshot Caption Gobbler
• Tiny Tools Beat Big Promises — the product principle behind the scope cut
• Ship the Experiment, Keep the Lesson — how to extract the learning after it ships

FAQ

Why cut features before building instead of building them first?

Because the cuts are much cheaper before the code exists. The Lie Detector makes the unsupported claims visible in one pass — before any implementation decisions are attached to them.

Should I always use Tarot as the first step?

No. The Tarot is useful for generating a starting constraint when the idea is still too open. If you already have a specific, scoped idea, skip it and go straight to the Lie Detector.

How many names should I generate before choosing?

At least 5. You need options that fail obviously so the good one becomes visible. One or two candidates is usually too few to see the contrast.

What if the Lie Detector flags everything in my idea?

That is useful information. It means the idea either needs to shrink to a smaller core, or the current brief is written in unverifiable language that can be replaced with specific statements. Start by listing what the tool actually does in one sentence without adverbs.