Why You Should Launch Products You're Ashamed Of

Most founders launch when they’re ready. Coco Mao, the CEO and co-founder of OpenArt — a 10-person AI startup with 6-7 million monthly active users and over $20M ARR — launches when she’s embarrassed. And she thinks more founders should do the same.

Her operating heuristic, repeated multiple times in our conversation: “You launch something that you are ashamed of and then you can test if this something that you are ashamed of can even attract some users or get good user feedback. And then you can polish from there.” The word “ashamed” is doing real work. Most founders interpret “ship the MVP” as “ship the polished MVP.” Coco’s version is structurally different — and the difference is the whole point.

The Polished Launch Is a Signal Trap

There’s a specific failure mode most founders don’t recognize. You spend three months making the product look great. You launch. Users come. They engage. You celebrate.

Then — six months in — you can’t tell why. Are users staying because the underlying value is real, or because the polish made it impossible to evaluate the value? You shipped a confounding variable.

Coco illustrates this with one of the more memorable analogies in startup advice:

“Imagine you are a girl, and you are all well dressed up and you attract a lot of attention. And then you would wonder, oh, is it because I’m just pretty or because people really like me? But then if you dress up very messy and then you go out and people still want to talk to you — those people are probably actually interested in talking to you. Instead of just looking at your look.”

Translated to product: a polished MVP can attract users for the wrong reasons. It can survive on visual gloss long enough that you mistake aesthetic appeal for functional pull. The dressed-down version strips that out. If users engage with something rough, they’re engaging with the underlying value, not with the wrapper.

What “Ashamed” Actually Means in Practice

For Coco, “ashamed” doesn’t mean broken. It means visibly incomplete. The version of the product that you’d hesitate to put on a Twitter announcement. The version that has rough UX, missing features, and clear gaps you intend to fix.

OpenArt’s earliest version was exactly this. In 2022, Coco and her co-founder spent less than a week crawling AI images from the web, pairing them with their prompts, and putting up a browsing site. They titled it “Pinterest for AI images” and posted it on Hacker News. There was no creation tool. There was barely any product. It went to number one on Hacker News that day.

That signal — that strangers actually wanted what they’d built — was what told them to keep going. If they’d spent three months polishing a Pinterest-for-AI-images product, the same launch would have produced the same number-one Hacker News post, but they wouldn’t have known if the polish or the underlying need was driving the result.

The point isn’t that products should stay rough forever. It’s that the embarrassment phase is when you find out whether you have something. After that, polish away.

The ICP Filter on Top

There’s a complication. Once you launch the rough version, you’ll get feedback from a lot of people. Most of them aren’t your customers. If you listen to all of them, you’ll polish toward the wrong target.

Coco’s solution is the ICP filter:

“You’ll have a wide range of audience. That’s why you need to be clear about your ICP, your ideal customer profile. So that you pick the problems of your ICP and then work on it instead of listen to everyone’s problems and then solve, try to solve everyone’s problems.”

OpenArt’s Discord has 350,000 users. Many of them have feature requests. Coco talked to 30 of them in a single week to figure out which ones were the actual ICP — gen-AI social media creators and SMBs, not casual hobbyists. Once you know who you’re building for, the embarrassed-launch signal becomes interpretable. Are the right users engaging? If yes, polish. If no, pivot.

The Speed Multiplier

The deepest reason this approach works is what it does to your iteration cycle. Coco’s broader thesis is that “making decision fast is more important than making the right decision,” because waiting at a fork doesn’t generate new information. The embarrassed launch is that thesis applied to product. Don’t sit on a decision about whether the feature works. Ship the rough version, watch what happens, and let the data tell you.

Most founders treat the launch moment as a final exam. Coco treats it as a probe. The difference compounds: if you launch every two weeks instead of every six months, you compress the learning cycle by 12x. Twelve times the experiments, twelve times the corrections, twelve times the chances to find the path that actually works.

The launches don’t have to be huge. Most of OpenArt’s launches aren’t. Coco notes the company actually under-markets — they ship continuously, and only do “fancy marketing videos” for big bets like the one-click story feature. The continuous shipping is where most of the learning happens.

FAQ

How do you know if your MVP is rough enough to launch?

If you’d be uncomfortable putting it on a Twitter announcement, it’s probably right. The discomfort is the signal that you haven’t over-polished. Coco’s version: launch when you’d be embarrassed to share it widely, but the core value should still be visible. If users can’t tell what you’re trying to do, it’s too rough — go back and clarify the core action.

What’s the difference between launching ugly and launching broken?

Ugly means the value works but the wrapper is rough — basic UX, missing features, no marketing. Broken means the value doesn’t work yet. Launch ugly, never broken. The point of the rough launch is to test demand for the core value, which means the core value has to actually function. OpenArt’s first version was ugly (just browsing, no creation) but the browsing experience worked.

How do I get feedback on a rough product without users dismissing it?

Pick a small, specific audience that’s predisposed to early products — Hacker News, niche Discord communities, beta lists, indie hacker forums. Skip mainstream channels (Twitter, Product Hunt) until later. Coco’s first launch was Hacker News specifically because that audience evaluates ideas, not polish. Mainstream audiences need polish to engage.

How do you decide what features to build after a rough launch?

Ignore feature requests. Coco’s principle: “You listen to the users, but you don’t listen to their feature requests. You listen to their problems.” Users will ask for solutions; what you want to know is the underlying problem. OpenArt’s users asked for “consistent characters.” The real problem turned out to be “they were trying to tell a story” — which became the visual storytelling pivot.

Will competitors copy a rough MVP?

Not as fast as you’d think. The signal of a rough launch is that most observers can’t tell whether you have something real. Competitors hesitate to copy because they can’t see the value clearly either. By the time they decide to invest, you’ve iterated three more cycles. OpenArt launched its Pinterest-for-AI-images version in 2022 — competitors arrived later, by which point OpenArt had already pivoted to creation tools.

When does the rough phase end?

When you can describe your ICP clearly and you’re consistently retaining users from that ICP. At that point, polish becomes worth the investment because you know who you’re polishing for. OpenArt’s rough phase lasted roughly through 2022. By 2024, they’d locked in image generation as the focus and polished aggressively. The shift wasn’t arbitrary — it followed the moment they could see who they were really building for.

Why does shipping ugly work in AI specifically?

Because the model is the value, and the model usually works. If the underlying AI is producing good output, users will tolerate rough UX to access it. AI products are particularly suited to embarrassed launches because the magic moment is in the generation, not in the wrapper. You can ship a button that triggers a workflow with no styling, and if the output is good, users will use it.

What’s the role of marketing in a rough launch?

Minimal. Coco’s view: “We probably don’t do enough formal launches” — most of OpenArt’s launches are continuous shipping with minor announcements. Save the big marketing for moments when you have something genuinely worth amplifying. Marketing a rough product can backfire by attracting non-ICP users whose feedback will distort your roadmap.

How long should you wait for data after launching ugly?

“You must look at least weeks of data, and then you do qualitative testing — basically talking to users who actually use the feature.” Days aren’t enough. Weeks let you see retention patterns, not just initial spikes. Coco’s standard: ship in one to two months, then watch for at least two to four weeks before deciding what to iterate on or pivot from.