Should I vibe code
Vector illustration and graphic design across Apple devices
A canvas with draggable shapes is a weekend. The pen tool is a career. Ask anyone who has shipped one.
?
Their verdict, the Pro price and the build-time estimate come from their entry, MIT-licensed. Checked 2026-08-04.
?
Our verdict, the regret score and everything below it. Editorial and unsponsored — nobody can pay to be moved.
The honest answer
why the verdict is what it is
Nothing here can hurt anybody, which is why the verdict is about the gap rather than the danger. An agent will hand you a canvas with draggable shapes, a colour picker and SVG export inside an afternoon, and it will look like eighty per cent of a vector editor. It is not; it is the eighty per cent that took a weekend, and the remaining twenty is where every vector application in history has spent its decade. The pen tool alone — handle mirroring, smooth-versus-corner nodes, snapping that guesses right, path simplification that does not destroy the curve — is a research project with a UI attached. Then boolean operations on overlapping paths, stroke-to-outline, text on a path, non-destructive effects, an undo stack that survives a hundred operations without drifting, and a file format that opens correctly next year. Curve does all of that with pressure-sensitive Apple Pencil input on hardware you cannot emulate in a browser. Build it for the pleasure of understanding béziers, which is a real pleasure. Do not build it expecting to stop paying for a design tool.
What actually breaks
not "if". the specific failures.
- The pen tool, immediately and permanently. Handle mirroring, corner-versus-smooth nodes and curvature-preserving simplification are the whole craft, and the first version will fight you on every stroke
- Boolean operations, which need robust path intersection with tolerances that behave sensibly on nearly-coincident curves — the classic case where a naive implementation returns geometry that renders but is not a shape
- Your file format, on the day you add a feature and last month's artwork opens with the layers flattened
- Undo, once effects, groups and transforms are in play and the stack starts recording states that no longer describe the document
- Text, in every direction at once: shaping, kerning, text on a path, and fonts that exist on your machine and nowhere else when the SVG travels
- Export fidelity — the thing you drew and the thing the printer receives disagreeing about colour space, stroke alignment or how a dashed line ends
- Performance, at about the fifth thousand path, which is roughly one detailed illustration
Is that you?
the verdict is a default, not a law
- You want to learn how béziers actually work, and shipping is not the goal
- The scope is one narrow job — a diagram generator, a logo mark placer, a batch SVG transformer — rather than an editor
- Everything saves as plain SVG that another application can open without your code existing
- Losing the whole project would cost you a weekend and nothing else
- Client work depends on it and a corrupted save is a missed delivery
- You need pressure-sensitive stylus input, which is where Curve's actual value lives and where a web canvas gives up
- Files have to open correctly in a year, in an application you have stopped maintaining
- You are hoping to cancel a design subscription rather than to learn something
If you build it anyway
the checklist, then the prompt that enforces it
- Save as plain SVG, not as a bespoke JSON document. A format only your app can read is how a year of artwork becomes unopenable, and SVG costs you almost nothing to adopt.
- Do not implement path geometry yourself. Use a maintained library — Paper.js, or a real boolean-operations library — and spend the time on the interaction instead.
- Autosave to a new file on a timer and keep the last ten. Every crash you have not had yet happens mid-drag.
- Pick one narrow thing and finish it. A great diagram tool beats a mediocre Illustrator by an enormous margin, and it is achievable.
- Write the export test early: render your output and the reference renderer's output and diff the images. Silent divergence is the bug you will not notice until print.
- Model undo as a command stack from the first commit. Retrofitting undo into a mutable document model is a rewrite, not a feature.
I want to build a vector illustration app. Keep me honest about the scope and
protect my artwork before you make anything look good.
1. Before any UI, tell me plainly which parts of a vector editor are multi-year
problems: the pen tool, boolean path operations, stroke-to-outline, text
shaping and on-path text, and non-destructive effects. Ask me which single
one I actually need.
2. The document format is plain SVG, read and written, from the first commit. If
I ask for a custom binary or JSON format, say no once and explain what
happens to last year's files when the schema changes.
3. Build save, autosave and export before the drawing tools. Autosave on a timer
to a new file, keep the last ten, and never overwrite in place during a drag.
4. Model the document with an explicit command stack so undo and redo exist
before the second tool does. Retrofitting undo is a rewrite.
5. Use a maintained geometry library for path intersection, offsetting and
boolean operations. Do not write curve-curve intersection yourself, and tell
me the tolerance behaviour you chose.
6. Add a rendering-diff test early: render a set of reference documents with my
renderer and with a known-good SVG renderer and compare images. This is how I
find out my strokes are misaligned before a printer does.
7. Implement the pen tool last, and when you do, implement node types and handle
mirroring properly rather than approximating them. A pen tool that fights the
user is worse than no pen tool.
8. Keep fonts out of the document: reference them, warn on export when a font is
not embeddable, and offer convert-to-outlines.
9. Watch performance from the start — tell me where this design stops working,
in paths, and make that number visible.
10. Out of scope, and say so instead of half-building it: pressure-sensitive
stylus input, colour management for print, collaborative editing, and
gradient meshes. If I need those, that is when a $30-a-month subscription
becomes the sane answer.That one keeps you out of trouble. For the prompt that actually builds it, canivibecodeit.com has one.
their build prompt ↗Or don’t build it
the boring option, and the way back out
The moment the work is for someone else. Thirty dollars an editor per month buys a pen tool refined over years, a file format that will still open, and print-grade export — none of which you can approximate. If the subscription is the objection rather than the tool, Inkscape is free, mature and opens the same files, and it is a far better answer than a vector editor you maintain alone.
$30/mo is cheaper than your weekend.
If the files are SVG, there is no exit problem at all — Inkscape, Illustrator, Figma and a browser all open them tomorrow, and your app becomes optional the moment you get bored of it. That single decision is the difference between a project you can walk away from and a folder of artwork held hostage by a parser you wrote. If you have already invented a format, write the converter now, while you still remember what the fields meant.
The mature free vector editor, with the pen tool, boolean operations and SVG fidelity that a homemade version spends years approaching.
Open-source design and prototyping platform that runs in the browser and stores work in open formats.
Embeddable browser-based SVG editor — the sensible starting point if you want an editor inside your own application.
Questions
Canvas libraries have made this easy. Hasn't the gap closed?
The rendering gap has closed almost entirely; the interaction gap has not moved. Drawing a bézier is a library call. Making a curve feel right when someone drags a handle — snapping, constraint, mirroring, tangent continuity, the difference between what the user asked for and what they meant — is the product, and no library ships it. That is why the demo is an afternoon and the tool is a decade.
What is the one decision that decides whether this was worth it?
The file format. Save plain SVG and your app is a nice thing you made that anyone can walk away from, artwork intact. Invent a format and every hour of drawing is locked behind a parser only you maintain, which is the one way a harmless weekend project can actually cost you something.
Could I build a useful subset rather than the whole editor?
Yes, and it is the recommendation. Narrow, opinionated vector tools — a diagram generator, a batch SVG transformer, an icon-set builder, a chart composer — are achievable, genuinely useful, and often better than the general-purpose tool at their one job. The failure mode is calling it a design app and then discovering how much a design app is.
Every week, someone ships something they shouldn’t have.
New verdicts, the worst thing that landed in the trap, and the occasional incident report. No other email, ever.
The editor is hard, the template library is harder, and the font licensing is a legal department.
Every operation is a library call. The product is that it runs in a browser tab on a 40MP file without dying.
Premium unlocks no features. It hides an ad. Rebuilding a decade of PSD archaeology to dodge a banner is the trade.
last reviewed 2026-08-05 · verdict is editorial and unsponsored · shared entry data from canivibecodeit under MIT · not legal advice