Melius Burned Its Whole Codebase. Then It Raised $25M. Should You Start Over?

"We scrapped the entire codebase; we burned all of it."

That's Joowon Kim, co-founder of Melius, talking to TechCrunch on October 6. He and two co-founders, all former engineers at Ramp, spent more than six months building an AI-powered performance marketing tool. Then they decided the idea didn't "have legs."

So they deleted it and started fresh. The new product generates ad campaigns, images, and videos. Melius says it passed $1 million in annualized revenue within two months of coming out of stealth in July. This week it announced $25 million in total funding: a $20 million Series A led by CRV and a $5 million seed led by General Catalyst.

Stories like this make starting over sound heroic. Most of the time, it's just painful. So how do you know when a restart is the right call and when you should keep fixing what you have? Let's look at what Melius did and turn it into a test you can run on your own product.

They kept the customer and changed the job. That's the most useful detail in the story.

They didn't jump to a totally new market. They stayed close to the people they already understood and moved to what they saw as a much larger problem those people had.

That pattern is common. Wilbur Labs' 2026 survey of 200 founders found that 81% pivoted from their original idea at least once, and 57% made a major pivot or several. Most good pivots aren't random leaps. They're a step sideways toward a bigger pain.

Founders far more often regret pivoting late than pivoting early.

In the same Wilbur Labs survey, 42% of founders said they wished they'd pivoted or changed their business model sooner. The researchers put it bluntly: startups rarely fail because founders change direction too often. They fail because they hold on to the wrong idea for too long.

Why does that happen? Sunk cost. Six months of code feels like an asset. Throwing it away feels like admitting those months were wasted.

But they weren't. Melius' team didn't lose what they learned about marketers, ad workflows, and their own ability to ship. They only lost the code. And in 2026, with AI coding tools, rebuilding code is far cheaper than it used to be. Learning is the expensive part, and that stays with you.

Score your product, without flattering it, against these five questions. One point for every "yes."

1. Are users polite but not pulling? They say it's interesting, then don't come back without reminders.
2. Have you shipped three or more major changes without a clear jump in usage? Iteration that doesn't move the numbers is a signal.
3. Do customers keep describing a different problem? If every call drifts toward a nearby pain, listen.
4. Is the code shaped around the wrong core idea? If the new direction would fight your data model at every step, patching costs more than rebuilding.
5. Would you build this product from scratch today if you hadn't already? This is the sunk-cost test.

0 to 1 points: keep iterating. You probably need sharper positioning or a better onboarding flow, not a new product.

2 to 3 points: run a small test of the new direction before touching the old code. A landing page, a manual service, or a prototype.

4 to 5 points: a restart deserves a serious look. Set a deadline, write down what you're keeping, and move.

Keep the learning, the customers, and the team. Let go of the code.

Here's a simple checklist for a clean restart:

• Write a one-page "what we learned" memo. What did customers say? What surprised you? Which assumptions broke?
• List every customer who gave you real time. Call them first about the new direction. They're your fastest feedback loop.
• Keep the boring pieces that work. Login, billing, and deployment scripts may survive even if the core product doesn't.
• Set a short clock. Give the new version a fixed window, say eight to twelve weeks, to show a clear signal.
• Tell your investors early. A clear explanation of why you're changing course builds more trust than a surprise later.

Before you commit, write down the new problem, the customer, and the first test in plain sentences. A planning tool like Foundra can help you lay out the new direction next to the old one, so the trade-offs are on paper instead of in your head.

A crowded market can be a good sign, as long as you know why you'll win a slice.

Melius isn't alone. TechCrunch lists Higgsfield, Krea, and Flora AI as competitors. Higgsfield was valued at $5.4 billion in August after raising $400 million, and TechCrunch reports it has passed $700 million in annualized revenue.

Kim's response was calm: "There are a lot of players, which is pretty exciting. That means there are customers to be won and there is demand in this space."

He has a point. Competition proves people will pay. But it also raises the bar. If you're restarting into a busy category, you need one clear reason a specific customer would pick you. Faster? Cheaper? Built for a niche the giants ignore?

Funding news from the same day shows how much money is flowing into AI products. Big rounds for others don't guarantee room for you, though. Your edge has to be specific.

Here's a made-up example to show why timing matters.

Imagine a two-person startup burning $25,000 a month with $300,000 in the bank. That's 12 months of runway.

• Restart at month 4: 8 months left. Enough for a rebuild, a launch, and a few months of traction before raising.
• Restart at month 8: 4 months left. Barely enough to rebuild, with no time to prove anything.
• Restart at month 10: 2 months left. You're raising on a story, not results.

The product decision is the same in all three cases. The difference is how much runway you have to make it work. That's why the 42% who wished they'd pivoted sooner matter so much.

Run this math on your own numbers today, and decide on the latest month you'd still have time to restart.

Tell it as a learning story with evidence, not as a confession.

Melius raised after its restart, not before. That order matters. The new product had early revenue to point to, and the founders could explain exactly why the first idea fell short.

If you're restarting and plan to raise, prepare three things:

1. The one-sentence reason. For example: "Marketers didn't need help managing ad spend nearly as much as they needed help making the ads."
2. The proof you learned something. Quotes from customer calls, usage charts from the old product, or the specific moment the pattern became clear.
3. The early signal from the new direction. Waitlist sign-ups, pilot customers, or first revenue. Even a small number beats a promise.

Avoid blaming the market, the timing, or your old investors. Own the call. Investors back founders who can see a problem early and act on it, and a clean restart is evidence that you can.

Also keep your existing backers in the loop before anything goes public. Nobody likes reading about a pivot in the press.

What is Melius? Melius is a New York-based AI platform for generating ad campaigns, images, and videos. It describes itself as an "agents lab for creative work."

How much did Melius raise? $25 million total, according to TechCrunch: a $20 million Series A led by CRV and a $5 million seed round led by General Catalyst.

Why did Melius scrap its first product? After more than six months of building an AI performance marketing tool, the founders decided the idea didn't have legs and moved to what they saw as a larger opportunity in creative generation.

How common are pivots? Very common. In Wilbur Labs' 2026 survey, 81% of founders said they pivoted at least once, and 42% wished they had done it sooner.

Should I delete my code when I pivot? Only if the old code is built around the wrong core idea. Keep reusable parts like login and billing, and always keep your notes on what customers told you.

1. Ex-Ramp engineers raise $20M for platform Melius after scrapping their first product, TechCrunch, October 6, 2026
2. Why Startups Fail (2026): Lessons From 200 Founders, Wilbur Labs, April 28, 2026
3. Higgsfield raises $400M Series B, quadrupling its valuation in 8 months to $5.4B, TechCrunch, August 17, 2026
4. Startup Funding News Today, October 6, 2026, Tech Startups