Most web converters upload your documents to someone else's cloud. Fine for a meme, not fine for a contract.
Starting point: "Convert to it!" by p2r3 (seen here). An in-browser converter with a great concept and a rough UI.
Conversions and PDF edits run entirely in your browser. Nothing to hack, because nothing gets sent.
No tracking. No analytics. No data sent home. There's a desktop helper for tricky formats, and it stays on your machine too.
"the only server involved is the one serving the page."
I took the rough project into an AI-build challenge. Claude Code wrote most of it; I directed: vision, design, tone. frogConvert also exposes an API, so other AI agents can drive it too.
Before building, I asked Claude to brainstorm the file structure, plan the architecture, and write it down.
That documentation becomes a shared reference, so Claude always has context instead of re-reading every file each time.
I asked Claude to write tests for every feature, then run them to verify its own changes.
You set the quality bar.
AI enforces it.
Unit tests and full browser tests, all AI-written.
creative director, not QA team
Use what's already out there. Browse GitHub repos, grab components from shadcn, lean on existing libraries.
AI is great at wiring things together. You don't need to build everything from scratch.
format conversions
"pdf to docx? attempting to undo what adobe hath wrought."
Four operations, all in your browser, built on proven open-source PDF libraries.
Mobile got the same polish: long-press to drag, a contextual action bar instead of fiddly per-card buttons.
PNG to PNG? It compresses instead of passing through unchanged. Long conversion? A live progress line keeps you in the loop.
No more dead-ends: every quality ceiling is an automatic degrade with a visible notice, not an error.
frogConvert is now an installable PWA. Works offline once handlers cache. Hooks into the OS share sheet and "Open with…" menu.
Close a tab mid-conversion and your work waits for you next time - files, target format, watermark settings, undo / redo, all of it.
Compression was always running - every conversion quietly applied a middling preset - but you could not see it or change it. It is now its own surface, with a level you pick, for images, audio, video and PDFs.
Compressing a PDF properly meant shipping a real PDF engine. That engine also reads PostScript, so a format family the app never supported came along with it.
Same engine under all three. Pick the surface that matches the task.
Eight tools total: three for conversion, one for compression, four for PDF editing.
Let's try it.
(- _ -) drop a file, pick a format
or edit a PDF
Start small. Pick a fun side project and build it with AI. That's how you learn.