Skip to main content

PortfolioCase study

TG-01Live pilot

Tegaki

Two pages. One human reader.

What your handwriting suggests about you — read and written by hand. A D2C pilot that turns a manual handwriting-analysis practice into a product a stranger can order.

Tegaki's live landing page: “Your handwriting has been talking this whole time.” on vermilion, beside an illustration of a reader holding a handwritten page.
  • 2 pages

    of handwriting, photographed or scanned

    Prototype

  • 1

    human reader — no AI in the product

    Structural

  • 3

    reading depths to choose from

    Prototype

    the pilot checkout confirms without charging

Evidence keyStructuralPrototype

The question

Would a stranger trust — and pay for — a reading by hand?

Tushar’s handwriting analysis was a fully manual practice: an analysis, then a polished report. The pilot asks whether that craft can become something a stranger orders, trusts and pays for.

The practice today
  1. A handwriting sample
  2. Analysed by hand
  3. A polished reportdocx / PDF

The product

You send. We read. You receive.

A buyer signs in, uploads two pages and picks a depth. A person reads the pages and writes a report that names what is visible before what it may suggest.

How it works, from the live site
  1. You sendtwo pages, photographed or scanned
  2. We reada human looks at the patterns
  3. You receivea personal report to reflect on
Tegaki page “Three things on a page, and what they suggest.”: a handwritten note beside “What we notice — the letters lean right” and “What it may suggest”.
What we notice, then what it may suggest
Tegaki “How it reads.” page with an excerpt from an illustrative sample report for a fictional subject.
A sample report (fictional subject)

The insight

The operator half is the product.

  • “The operator half is unspecified (biggest gap)… Turnaround promises live or die in… the admin surface.”

    — Tegaki Discovery PRD §5.1

Product decisions

A small, honest pilot before a big launch.

  • Could have
    Generate readings with AI
    Chose
    Keep the reading human
    Because
    The pilot asks whether a stranger would trust this experience; report writing stays manual and offline.
  • Could have
    Launch with payments, a domain and email
    Chose
    Pilot first: checkout confirms, charges nothing
    Because
    The first goal is 5–10 pilot users completing the whole loop.
  • Could have
    Promise answers about who you are
    Chose
    Indicative readings, guarded in CI
    Because
    “Readings are indicative and growth-oriented, never a diagnosis” — and a CI honesty check guards that line.

How it works

Private by construction.

  1. Google sign-in
  2. Upload two pagesbuyer-scoped storage
  3. Row-level security18 database migrations
  4. A human reads
  5. Report deliveredsigned URL
  6. Retention jobscheduled clean-up
A handwriting sample’s path through Tegaki

Evidence

Live and guarded — still waiting for its first strangers.

  • 3

    layers enforce the sample rules

    Structural

    browser, server action and database CHECK constraints

  • 18

    migrations with row-level security

    Structural

What isn’t measured

  • No pilot users or orders are recorded yet.
  • The checkout confirms an order without charging for it.
  • The 5–10-user pilot goal hasn’t been measured.

Sources

Everything here is backed by a real artifact.

  • PRD
  • Readme
  • Code
  • Live data

Evidence behind the Tegaki case study

Private working documents are listed with what they support; only public sources link out.

  1. Discovery PRD

    PRD1 Sep 2026

    Supports: The manual practice, the buyers and the operator-half gap

  2. Solution PRD

    PRD1 Sep 2026

    Supports: The pilot question and the 5–10-user goal

  3. README

    Readme

    Supports: No AI in product; guardrails in three layers; never a diagnosis

  4. Build: package, migrations, tests

    Code

    Supports: 18 RLS migrations, 31 test files, the check-claims guard

  5. Live pilot pages (opens in a new tab)

    Live data

    Supports: How it works, the anatomy page and the sample report

Links and next steps

Solo buildSep 2026Live pilot