Windsurf vs Cursor: what shipped, what earned

Cursor is named in 44 revenue-graded cases in the ProvenStartups index. Windsurf is named in 2, and one of those two is Windsurf's own company record. Treat the Windsurf column here as thin evidence, not a verdict: two cases cannot rank an editor, they can only report what got published.

Contents

• ·Which tool's builders show more revenue evidence?
• ·What do the top cases on each side actually earn?
• ·Where do the cases on each side cluster?
• ·What do the evidence grades say about each side?
• ·Which one should you pick?
• ·What can't these numbers tell you?
• ·Frequently asked questions

Which tool's builders show more revenue evidence?

Cursor, by a wide margin: 44 graded cases name Cursor against 2 that name Windsurf. That counts published breakdowns, not code quality. Cursor arrived earlier in the creator economy that produces these interviews and became the tool founders say out loud on camera. Evidence volume tracks attention, and attention is not causation.

The shape of the asymmetry matters more than the ratio. The Windsurf side holds Windsurf (Codeium) itself — a $1.3B valuation, 1M+ developers, roughly 200 staff — plus one operating business built by someone else. A vendor's balance sheet is not evidence that its users earn. The same applies to Anysphere: $500M/yr and a $9B valuation prove the editor sells, not that it makes your app sell.

So the real comparison is not 44 versus 2. It is what the 44 share, and whether the Windsurf pair contradicts any of it. Nothing does, which is all a sample that small can say.

What do the top cases on each side actually earn?

The top verified Cursor case is Gravl at $440K/mo with more than 70,000 subscribers. The top Windsurf case that is an actual product business is Cluely at $500K/mo, roughly $6M ARR, founder-reported. Both are outliers. Across Cursor cases publishing a clean monthly figure, the median sits near $14K/mo.

Grades, first use: ✅ third-party verified, 🗣 founder-reported, 📎 creator-relayed, 🔮 unproven. The grade is not an accusation. It records how much support sits under the number, so Cluely's $500K/mo 🗣 and Gravl's $440K/mo ✅ never get compared as if they were the same claim.

Where do the cases on each side cluster?

Cursor's 44 cases spread across ten categories, and 27 of them are solo-run. Windsurf's two are both company-scale: Cluely runs 13–14 people plus about 60 contracted creators, and Windsurf itself has roughly 200. With two cases, that pattern says nothing about whether one person can ship with Windsurf.

The Cursor category split is the interesting part: 15 SaaS, 9 consumer apps, 5 scale references, 4 AI websites, 4 platform plugins, 3 directory sites, and one each of ecosystem tool, AI service, simple tool, and cautionary tale. By tier, 16 are Tier 1 (copy this now), 22 Tier 2, and 6 Tier 3.

Notice what is missing: one simple tool in 44 cases. The earners are ordinary software with a named buyer — a Shopify plugin, a Notion add-on, an Upwork bidding agent, a gym app. Our favourite detail sits in Cursor Directory: a three-hour weekend project aggregating Cursor rules, at $34,000–35,000/mo ✅. The best-verified mid-size earner in the cohort is a directory about Cursor, not an app Cursor wrote.

What do the evidence grades say about each side?

Of the 44 Cursor cases, 10 are ✅ third-party verified, 26 🗣 founder-reported, 4 📎 creator-relayed, and 4 🔮 unproven. The Windsurf pair is one verified (the vendor) and one founder-reported. So roughly a quarter of the Cursor cohort carries outside confirmation — better than most tool roundups, and still a minority.

Verification changes what a case teaches. Profit AI is graded ✅ because the founder read his Shopify partner dashboard on camera: $147,000 total since a December launch, 122 app users with 38–39 paying, and 47 installs against 43 uninstalls in the same window. That last pair is what no founder volunteers in a tweet. A verified figure carrying its own churn beats a larger unverified one.

It also explains our scepticism about the headline. The one ✅ on the Windsurf side is a valuation — the easiest number in our AI coding tools library to verify, and the least useful to copy.

Which one should you pick?

Pick on workflow fit, then on price, and ignore the case counts — they measure publicity. Nothing in either cohort shows either editor producing revenue the other could not. Both sides cluster around the same three ingredients: a buyer you can name, a channel you already have, and a scope small enough to finish.

You are a solo founder shipping your first paid product

Either editor works; the risk is not the editor. Of the Cursor cases, 27 of 44 are solo-run, which proves feasibility and nothing else. Choose the tool you will actually sit in for eight hours, then spend the saved time on the channel. See what is Cursor AI for the workflow specifics.

You are working inside a large existing codebase

Lean toward whichever tool holds more repository context in the way your project is laid out, and test both on the same real ticket for a week. Our Windsurf evidence review makes the same point from the other direction: Windsurf's adoption figures are strong while its per-app revenue evidence is almost absent.

You are standardizing a team on one tool

Buy on vendor durability and review cost, not on case counts. Both vendors are well capitalized: Anysphere at $500M/yr and a $9B valuation ✅, Windsurf at a $1.3B valuation with 1M+ developers ✅. Neither is a coin flip. Run a paid pilot on your own code and measure review hours, not demo speed.

What can't these numbers tell you?

They cannot tell you a tool caused anything. We index businesses whose founders published numbers, almost always in a video breakdown, and we record whichever tool they happened to name. That is selection bias twice: on who agrees to talk, and on what they remember to credit while talking.

Look at what actually moved each case. Cluely's engine was roughly 60 contracted creators pumping short-form video, not an IDE. Gravl copied the market leader's interface and replaced the training engine. Prayer Lock took three days to write; the real work was finding one video format that converted. Stoppr cloned a rival screen for screen and reached $12,000/mo 🗣.

In every one of those, the editor was a typing accelerator. The Profit AI founder describes the build as uploading a CSV, telling Cursor to build it, then finding all the ways Cursor got it wrong. Swap the tool name and the business is unchanged. Our full ranking is AI coding tools by revenue evidence; the cohort write-up is apps built with Cursor.

More in AI coding tools.

Frequently asked questions

Does 44 cases versus 2 mean Cursor is the better editor?

No. It means Cursor gets named far more often in the public breakdowns we index. Case volume measures publicity and timing, not compile quality, context handling, or cost. Two cases cannot support any ranking claim, in either direction. Judge the editors on a week of your own work, then use the case data to judge business shapes.

Why does ProvenStartups only have two Windsurf cases?

Because we only index a business when a founder publishes revenue figures and names the tool. Windsurf's own record accounts for one of the two; the other is Cluely, founder-reported. Adoption is clearly much wider than two projects. The gap is in disclosure, not in usage, and we would rather show two honest rows than pad the column.

What does a verified grade actually require?

An outside source or a figure read from a live dashboard on camera, not a founder's recollection. Profit AI qualified because the Shopify partner dashboard was shown, including uninstall counts that cut against the headline. Founder-reported and creator-relayed grades stay useful for pattern-finding; they just do not settle arguments about how much a business really earns.

Does the tool I pick change how much I can earn?

Far less than founders hope. Across these cases the earners share a named buyer, an existing channel, and a scope small enough to finish, and they range from roughly $700/mo to $500K/mo using whichever editor was handy. The tool changes how fast you produce a wrong version. Distribution decides whether anyone pays for the right one.