Loading…
Real milestones from building a PDF editor from scratch. Not version numbers — what actually shipped and why it mattered.
The visual editor shipped. Upload a PDF, AI converts it, edit with full reflow, export. Or create documents from scratch with templates. Free plan — 50 exports/month, up to 10 pages/doc, 5 AI conversion pages/month. Founding member pricing locked at $8/mo (annual) for early users.
When a flex container crosses a page boundary, splitting it changes the available space, which changes the flex distribution, which moves the split point. Every library I studied treats pagination as a separate step after layout. That's why they all break on multi-page flex. I had to solve them in the same pass. The implementation went through four rewrites before the edge cases stopped popping up.
Colspan, rowspan, automatic header repetition on every page. When a table overflows to the next page, the header row repeats and content continues. This is the single most requested feature I saw in issue trackers of other libraries — and the one most of them don't handle correctly.
TTF subsetting with glyph collection, composite glyph resolution, CID encoding, and ToUnicode CMap. Without that last part, copying text from a subsetted PDF gives you garbage characters. The glyph ID mapping had an off-by-one error that took three days to find. The fix was one line.
The implementation covers W3C CSS Flexible Box Layout §9 — main axis sizing, cross axis alignment, grow/shrink distribution, min-size constraints, frozen items, multi-line wrapping. Ran it against extracted Web Platform Tests. Standard patterns — card grids, header-content-footer, sidebar layouts — render correctly. Some obscure edge cases with percentage-based flex-basis still need work.
The engine generated its first valid PDF. Colored rectangles, Helvetica text, correct cross-reference table. Every PDF viewer opened it without errors. From this point every layout feature was visually verifiable — I could see the output instead of guessing from coordinate logs.
Input sanitization, URL validation with SSRF prevention, font magic-byte checking, image decompression bomb detection, resource limits. I studied every CVE in existing PDF libraries and built protections for each attack vector. All active by default, all before the first layout function existed.
Upload a PDF or create from scratch. Everything on this page is live and working.