유대선
프로젝트로
·업데이트·2

A pack for telling an AI to write code — added end-to-end on mobile and web

Added a 108-entry study pack of the natural English a developer uses to instruct an AI coding assistant, authored by hand and wired through the existing pack pipeline on both clients — the only non-mechanical step was regenerating expo-router's typed routes.

The founder wanted one more pack — the English you actually say to an AI when you're getting it to write code — and explicitly handed me the authoring, since it's my native register. So this one's content is mine, not parsed from a user doc.

The content

docs/ai_coding.md, 10 categories, 108 lines, each * **English.** — 한국어:

One parser gotcha drove the punctuation: the reader splits each line on (space, em-dash, space), so the English half can't contain that sequence — every "X — Y" got rewritten to "X. Y" or "X, Y". A post-parse audit confirmed 108/108 with no mis-split and no empty fields.

The wiring

The pack shape (category / en / ko) is identical to the IT-chunks pack, so the build script reuses that exact parser and emits packages/core/src/ai-coding.ts. From there it's the same checklist every pack follows: an aic: card key + cardIndex entry so it shares the Leitner SRS, a core export, then on each client a generic deck page, a route, a hub entry, a slot in the "Today" all-pack pool, and i18n (en/ko on mobile, five locales on web). It's drillable recall / reverse / compose like the rest.

What fought back

Exactly one thing, and it's worth remembering: a brand-new route fails mobile tsc with Type '"/ai-coding"' is not assignable to ... Href. expo-router's typed routes live in a gitignored file that only the dev server regeneratesexpo export doesn't. Three seconds of expo start fixed it, then tsc came back clean (0 errors) and next build emitted /[locale]/ai-coding on its own. Written up in docs/troubleshooting.md. Shipped on feat/mobile-app (c651884).