πŸͺΆ β€”
INKWELL

EXTRACT

Drop a file. Paste a URL. Snap a photo. Get clean text.
0 chars 0 words

HOW IT WORKS

FIND Local regex pass β€” strips HTML tags, decodes entities, kills [1] [2] citations, normalizes whitespace. Free, instant, never blocks. Best for tidying up your own pastes.

FIX Haiku 4.5 reads the mess and rebuilds it β€” strips nav cruft, ads, footers, share buttons. Preserves every word the author wrote. Best for messy article copies. ~1 ink per 10K chars.

FINISH Sonnet 4.6 polishes prose flow β€” fixes awkward sentences, kills redundancy, smooths transitions. Use only on your own drafts. Will not summarize. ~5 ink per 10K chars.

SORT Sonnet 4.6 reads tabular or list data, auto-discovers natural categories, returns an organized markdown report. Drop a spreadsheet, get sense of it. ~4 ink per 100 rows.

READ IMAGE Sonnet 4.6 vision transcribes handwritten setup sheets, pit notebooks, index cards, photos of any document. Preserves tables, label/value pairs, lists. Crushes traditional OCR on cursive and scribbled numbers. Flat 2 ink per image.

Files dropped: text and markdown load locally. Spreadsheets (XLSX, CSV) parse client-side. PDFs extract text via pdf.js (digital PDFs only). Photos route to Sonnet vision (resized to 1568px before upload). Audio and video route to Whisper transcription.

URL pasted: auto-fetches via Readability β€” same algorithm Firefox Reader Mode uses. No LLM, no cost.