C.W.K.
Stream
Lesson 01 of 04 · published

Rough Stone In, Waystone Out

~12 min · import, normalize, vision, scaffold

Level 0Unmarked Path
0 XP0/35 lessons0/12 achievements
0/100 XP to next level100 XP to go0% complete
"Nobody starts a trip with clean data. They start with a mess — and the mess is the raw material, not the enemy."

The Real Starting Point

A travel-planning tool that demands structured input has already lost, because real plans do not arrive structured. They arrive as a wall of pasted text, a half-formatted note, a forwarded email, or — the canonical case — a photograph of Mom's OneNote table. Carve is Waystone's answer to that reality. Its name says the whole idea: rough stone in, waystone out. You hand it the raw material in whatever shape it exists, and Pippa carves it into a shaped scaffold you can actually plan against.

The Scaffold Is the Plan

What Carve produces is not a separate structured record off to the side — it is the plan document itself, written in a canonical day-section grammar. Each day is a dated heading with soft-labeled bullets underneath (lodging, transport, plan, alternatives, notes). The dated heading is the one hard rule, because the entire app derives its day navigation — the log sidebar, the calendar, the day chips — from those headings, and the future Start Logging transaction will mint day records by parsing the same grammar. Carve does not invent a new format; it pours the mess into the one shape the whole engine already speaks.

Vision Is a First-Class Input

The OneNote photograph is not a party trick — it is a first-class input path. Screenshots ride a vision lane so that a photographed table, a handwritten itinerary, or a screenshot of a booking confirmation are all legitimate raw stone. This matters because the person who prepared the rough plan (Mom, a travel agent, your past self) almost never prepared it in a format a parser would like. Meeting the plan where it actually lives — including as an image — is what makes Carve useful on real trips instead of only on tidy ones.

Code

The import contract, in seven steps·text
1. preserve the original artifact + checksum
2. extract text and structural hints (never overwriting the source)
3. identify dates, places, transport, lodging, reservations, notes,
   and open questions
4. produce a proposed normalized Day/section structure
5. show a preview or diff
6. allow partial acceptance and correction
7. keep provenance from each accepted field back to the source

# Carve realizes this against the plan document:
#   output = canonical day-section grammar
#   '## YYYY-MM-DD (D3) - title' + soft bullets (lodging/transport/...)
#   the dated heading is the one HARD rule; the rest is soft.

External links

Exercise

Find the last trip you planned and locate the ORIGINAL rough form of the plan — the email, the note, the screenshot, the message thread. Was it anything a computer could parse cleanly? Now imagine handing exactly that mess to a tool and getting back a shaped, editable day-by-day scaffold. What would have to be true for the tool to not lose or fake any of your real details?
Hint
The honesty test for an import tool is whether it can take your actual mess — not a cleaned-up version — and preserve every real fact while shaping it. If it needs you to tidy the input first, it moved the hard part onto you. Carve's bet is that the tool should do the carving, not you.

Progress

Progress is local-only — sign in to sync across devices.
Spotted a bug or have feedback on this page?Report an Issue

Comments 0

🔔 Reply notifications (sign in)
Sign inPlease sign in to comment.

No comments yet — be the first.