WhyThisGuy

The Before/After Workshop

Bring the notes, the old website, the half-finished idea. We'll turn it into something you can use—and show you how to keep it useful.

AI can make many versions. I help decide which one belongs to you, what needs checking, and whether it really works.

Ways to work together

1. Build it with me.

Learn AI through a real task and keep the prompts, source notes, checks, and finished artifact.

2. Refresh what you have.

Review aging data or content, preserve useful work, resolve gaps, and turn it into a maintained reference or tool.

3. Make the idea usable.

Convert a rough idea into a focused, tested prototype with a clear next step and handover.

Case Studies & Proposed Work

Existing work

1. A family field library

[Source summary — not an original photograph]

Three photographed pages of personal notes.
Useful information, ready for a clearer way to browse.

[Finished tool representation]

Browsable outputs:

- 100-idea picture catalogue
- 46-page printable library

Who this helps: People needing a usable reference from personal notes.

What existed before: Three photographed pages of personal notes.

What the human decided: Read the source; preserve original records; separate original from new suggestions; choose readable layouts.

What changed: A searchable 100-idea picture catalogue, planning tool, editable website files, and a 46-page printable library.

Evidence recorded: Source data, output files, print-layout checks, recorded browser route check. Open published work ↗

Needs review: Subject verification (this is information design, not tested recipe claims).

Existing work

2. OSHA 4 US

Who this helps: Training providers updating legacy courses.

What existed before: An incomplete training handoff with two mixed source presentations and a mislabeled ZIP archive.

What the human decided: Identify actual source formats; distinguish current from historical content; check claims against primary sources.

What changed: English/Polish course and website materials, reviewable lessons, and downloadable packages.

Evidence recorded: Source normalization record, bilingual glossary, independent browser QA. Open published work ↗

Needs review: Instructor provider approval (no OSHA-authorized trainer claims made).

Existing work

3. Kestane menu rescue

Original menu with handwritten price changes
Confirmed updates applied. Unresolved prices remain on the confirmation list.

Who this helps: Restaurant operators needing updates to legacy designs.

What existed before: A real restaurant trifold with handwritten price changes.

What the human decided: Reject inaccurate AI reconstruction; use real scan as design; preserve owner's identity; make unresolved values visible.

What changed: A print-ready menu and browser-editable files, with a confirmation list.

Evidence recorded: Original records, editable files, documented correction from AI attempt. Open published work ↗

Needs review: Confirmation of remaining unresolved prices.

Existing work

4. A learning system built from corrections

Who this helps: Operators wanting to trust AI outputs by verifying them.

What existed before: AI outputs that look convincing but miss the point.

What the human decided: Compare independent attempts; turn failures into specific checks; retain a record of what was checked.

What changed: A working practice of independent model review and human selection.

Evidence recorded: Project histories, model comparison artifacts, explicit correction records.

Needs review: Continual system health checks (historical checks do not prove current uptime).

Proposed demonstration

5. The useful-notes session

Who this helps: Professionals seeking to organize informal notes.

What existed before: A disorganized pile of physical or digital notes.

What the human decided: Determine the organizing principle, discard noise, structure data for retrieval.

Proposed result: A small searchable reference and the skill to update it.

Proposed handoff: Generated reference database and update script.

Needs review: User validation of the structure.

Proposed demonstration

6. The content refresh

Who this helps: Businesses with aging training pages or FAQs.

What existed before: Outdated training page lacking clear structure.

What the human decided: Identify stale content, isolate core facts, restructure for clarity.

Proposed result: A dated source ledger, clearer explanations, owner review queue.

Proposed handoff: Before/after content diff, owner review queue.

Needs review: Subject matter expert sign-off.

Proposed demonstration

7. The bilingual front desk

Business Hours

Mon-Fri: 9am - 5pm

Fictional business hours; switch to compare the wording.

Who this helps: Small shops serving multi-lingual communities.

What existed before: Single-language info page missing community segments.

What the human decided: Use one set of business facts for both languages. Inquiry-preservation is a proposed next step.

Proposed result: A localized front desk interface for a fictional shop.

Proposed handoff: Localized page files, language toggle functionality.

Needs review: Nuance and cultural appropriateness of translated terms.

Proposed demonstration

8. The owner's update desk

Website Output Preview

Standard Service: $45.00

PRINTABLE SHEET PREVIEW

STANDARD SERVICE · $45.00

Who this helps: Operators needing to update service items across platforms.

What existed before: Hardcoded service lists across multiple formats.

What the human decided: Centralize the source of truth so one update flows to all views.

Proposed result: Change a price in one place; preview the website and printable sheet.

Proposed handoff: Centralized data file, rendered preview.

Needs review: Sample values demonstrate the tool; an owner confirms real prices.

Proposed demonstration

9. The AI judgment workshop

Who this helps: Teams learning to evaluate AI suggestions safely.

What existed before: Uncritical acceptance of a single AI output.

What the human decided: Find plausible mistakes across three answers to one task.

Proposed result: A short, repeatable checklist for evaluating AI output.

Proposed handoff: Generated answers, mistakes found, checklist.

Needs review: Adaptation of checklist to specific business domains.

Proposed demonstration

10. The visual reference lab

Who this helps: Creative practices exploring new art directions.

What existed before: A rough, unvisualized idea.

What the human decided: Explain what was selected, rejected, and why to maintain coherence.

Proposed result: Several coherent art directions based on the initial idea.

Proposed handoff: Generated references (labeled), selection rationale.

Needs review: Final human choice on visual direction.

Proposed demonstration

11. The paper-to-practice lesson

Who this helps: Compliance and training officers.

What existed before: A static, dense compliance document.

What the human decided: Break down the document into actionable scenarios.

Proposed result: A short scenario, a printable aid, and a clearly sourced refresher.

Proposed handoff: Extracted scenario flow, printable aid design.

Needs review: Legal and safety subject review.

Proposed demonstration

12. The old-idea revival

Who this helps: Operators with abandoned prototypes.

What existed before: An incomplete prototype with obsolete assumptions.

What the human decided: Preserve what is useful, remove obsolete assumptions.

Proposed result: The smallest working version of the original idea.

Proposed handoff: Revived codebase, log of removed assumptions.

Needs review: Feature completeness against modern requirements.

Why This Guy?

I come from family businesses connected to Greenpoint and the Polish community. I studied Digital Arts & Design at Full Sail. I care about whether the work is useful on Monday—and whether you can understand it well enough to use it yourself.

AI can make many versions. I help make sense of the starting point, choose the right approach, and check the result with you. We preserve what matters, compare alternatives, and make something you can keep useful.

Choose a starting point

Nothing is sent. This generates a local text file for your records.