What I learned shutting down CarbonSync

Earlier this year we shut down CarbonSync. My co-founder Keerthana wrote the company’s public wind-down statement, and you should read hers first. She named the real problem: the people who decide which materials go into a product mostly sit outside the design team, and the change we were chasing needs a bigger ecosystem shift than one tool can drive.

This is the operator’s side of that story. The part I actually learned, not the part we pitched.

CarbonSync was a GenAI tool for eco-design. The idea was to let a designer treat sustainability as a real design parameter, the same way they treat cost or weight, instead of a number someone reconstructs months later in a spreadsheet.

I didn’t set out to found a company. The honest version is messier than the founder story usually goes. After HealthifyMe, I had interviews lined up almost immediately. A few conversations in, I realised I didn’t want another job. I wanted to build something of my own. So I started consulting for startups, helping them with hiring and setting up teams. A mutual connection introduced me to CarbonSync because they needed help. I came in as a consultant, liked the problem enough to stay, and it became mine.

The problem was the right size. Big enough to matter, small enough that I believed I could solve it. And if I could make my corner of the world slightly better while making a living, that was a good trade.

We didn’t fail at the product. We failed at timing. That distinction is the whole post.

What worked

I want to be honest about this part, because it’s the part that still stings.

We made the bad data work. Anyone who has touched sustainability data knows it’s messy, incomplete, and inconsistent across sources. We built things on top of it that gave people real answers.

We built products people liked. These weren’t polite nods. Real feedback, asks for more, people who wanted it better instead of wanting it gone. People wanted the thing to exist.

We moved fast. We ran experiments quickly, killed what didn’t work, kept what did. We didn’t fail because we didn’t try. We tried hard, and we were good at it.

So if the product was liked and the team executed, what killed it?

What broke

We were building something futuristic, and we knew it. The whole bet rested on one assumption. EU regulation was coming, manufacturers knew it was coming, and they’d start solving for it early because they could see it heading their way.

That’s not how it went. The market was busy worrying about other things, political and economic noise that felt more urgent. Sustainability was the last thing on most buyers’ minds. Knowing a problem is coming, and putting budget against it today, are two completely different things, and we’d bet the company on them being the same.

And the tailwind we were counting on actually got weaker while we built. In 2025 the EU pushed its big sustainability reporting rules out to 2028 and watered them down under a simplification package. When the regulator itself blinks on the timeline, you have your answer about how urgent the market finds you.

You could see it in how we tried to sell. We started with prosumers and design studios. They liked it, but the subscription was small and the scale was never going to add up. So we made the product enterprise ready and went after bigger checks. Same result, one level up. Champions who genuinely loved it, and couldn’t get a yes.

Here is the number I can’t unsee. We were talking to a European enterprise doing around 500 million dollars in revenue. Their entire sustainability team was two people. Two. At that size, in Europe, where the regulation was supposed to be the strongest tailwind we had. That was the moment “wrong time” stopped being a worry and became a fact. If even they didn’t have the budget or the people, the buyer we needed didn’t exist yet.

By then I’d been at it a year, my co-founder longer. We looked at what it would actually take to crack this, and the answer was a long runway. In this economy, betting on a long runway to wait for a market to mature isn’t brave, it’s just expensive. So we made the call to wind down. Not because it was wrong. Because it was early.

The thing I stopped believing

I disagree with the common view that a problem everyone agrees is coming is a problem you can build a business on today.

Every investor, every customer, every regulator agreed sustainability mattered and that the rules were coming. Agreement was never our problem. We had all the agreement in the world and no budget to go with it. Awareness is not demand. A regulation on the horizon creates conversations, not line items. The 500 million dollar company with two sustainability people wasn’t confused about whether the problem was real. They just weren’t going to pay for it this year, or likely the next.

For engineers this one is worth sitting with, because we’re the people most likely to get it wrong. We believe that if you build something good enough, the market shows up. We obsess over the product because it’s the part we control. But the product was the easy part. We made bad data useful, shipped fast, and people liked it, and none of that could manufacture a buyer who had budget. Distribution and timing were the whole game, and they’re the part most of us treat as someone else’s job.

What I carry forward

The first thing I’d validate now, before building anything, is whether it’s a problem of today or a problem of tomorrow. Not “do you agree this matters,” because everyone agrees everything matters. The real question is “do you have budget for this, this year, and who signs the check.” We never pushed hard enough on that question early, because the answer might have stopped us, and we wanted to build.

But here’s where I keep arguing with myself, and I will not pretend I have it resolved.

The clean lesson is “solve a problem people have today.” And I mostly believe it. But with AI moving this fast, the safest-looking problem of today can be commoditised in a year, and some of the biggest opportunities are bets on a near future that doesn’t quite exist yet. So the lesson isn’t “never build for the future.” It’s harder than that. It’s learning to tell the difference between early and wrong, and refusing to bet a whole company on a future arriving inside your runway unless you have something that pays the bills while you wait.

I still think CarbonSync was early, not wrong. I still believe that product has a place, maybe a few years out, if the budgets and the teams ever catch up. Like Keerthana wrote, that’s a bigger ecosystem shift, not just a calendar problem. I just will not be the one funding the wait.

Close

I am not going to pretend there is a tidy moral here that makes it feel better. We built something good, people wanted it, and the market wasn’t ready to pay for it yet. That is allowed to just be disappointing.

What I took from it is sharper instincts about timing, distribution, and the difference between a problem people admire and a problem people fund. If you’re building in sustainability, or any regulated, ahead-of-its-time space, or you’re wrestling with that same today-versus-tomorrow call on an AI product, I’d like to compare notes. I’m at hello@balavishnuvj.com.