Ending Osmo

Last September, Arafat and I made a decision. We’d wrap up our current projects and, starting from October 1st, go all in on a startup. We didn’t have a perfect plan — just a blurry vision of what we wanted to build. Looking back, what we were really trying to build was a better Canva. A more opinionated, brand-safe, focused version.

I always appreciated how Canva gave non-designers the ability to make passable designs. But I hated how everything made on Canva looked like it was made on Canva. Generic. Same-ish. Instantly recognizable in a bad way. Arafat’s background in branding gave us an idea: what if design tools only showed you a limited, brand-specific set of options? Every brand would have their own curated elements, colors, layouts — Osmo would be the tool that enforced this identity, not diluted it.

We started off by trying to understand what features would be necessary to allow creation of social media posts. We quickly built a demo to show it to the folks at the Attention Network and a few others in our (mostly Arafat's) immediate network. We got both positive and negative feedback. However, after one of these conversations, a well-wisher gave us an advice that stuck with me:

Lesson 1: Don’t over-index on feedback. Everyone has opinions. If they’re in your target audience, sure — take notes. But if they’re not? Hear them out and move on. Early-stage startups should listen to usage more than opinions.

So we decided to focus on building and launching, and would only feel validated (or invalidated) by how people actually used it. During the demos, a tiny feature we added last minute — background colors adapting to the foreground — got people really excited. We leaned into it, branded it as “interactions,” and promised more such interactions would be coming soon to make creating designs so much easier. We thought this would help set us apart from Canva.

We were getting meetings before Osmo was even fully functional — a huge testament to Arafat’s credibility and network. That taught me:

Lesson 2: Founder-led sales is a cheat code. You get extremely fast results. You get to use the founder's credentials to get users. This is also why you should try to have at least one domain expert with the domain network on your founding team.

We onboarded Attention, set up analytics… and then saw something strange. They only used Osmo for adding the background and logos. Then they’d download the image and upload them to Canva. I tried to figure out why they were doing that. What I found was:

1. Osmo was too slow (because of some hacky stuff we’d done early on).

2. It wasn’t a self-contained experience — people had to keep switching tabs to find assets.

Both were solvable. But we were stretched thin. The MVP was messy. I’d hardcoded a bunch of things for the Attention launch that made onboarding others a nightmare. I was stuck between building new features vs generalizing the existing ones. It felt overwhelming. And it was slowing us down. We had a list of clients who were ready to use our product but we couldn't onboard them.

As if that wasn't enough, while we were demo-ing the product with Attention assets, one of our potential client didn't realise it due to a miscommunication and ended up downloading a post made using the brand elements of Attention Network and started running ads on Facebook with it. That ended up being extremely messy. It took us days to make sure all trace of that post was removed from Facebook. We decided to stop demo-ing others with Attention's assets and I took a big lesson here:

Lesson 3: Build an MVP that is demo-able and onboard-able without writing any additional code or wearing the CTO hat. This also means the non-technical founder should be able to both demo the product and onboard new users without bothering the technical founder.

Meanwhile, reality hit. We were running out of savings. Our parents were covering the essentials, but little things — hangouts, Uber rides to meetings, random coffees during work sessions — those started becoming harder to justify. Osmo wasn’t making money yet, and fundraising felt far off.

We applied to YC. Filled the form with a weird mix of hope and disbelief. Deep down, I knew we weren’t ready. We weren’t even full-time. Still in uni, not willing to drop out. YC rejected us (we weren't even in the top 10%). That stung, but it made sense.

So Arafat took on a few freelance gigs to stay afloat. That meant fewer check-ins. Fewer whiteboard sessions. Less energy. I started losing focus, and eventually the motivation. This hints at a very important lesson:

Lesson 4: Have a personal runway for 8–12 months. It buys you clarity, patience, and peace. You can’t create while trying to survive. I can imagine this being even more true if you're no longer a student or have a family to feed. Only exception to this is if you can make the startup fund you, through early revenue or fast fundraising.

A few weeks ago, Arafat and I sat down to talk about Osmo. He told me he wasn’t confident yet in his ability to generalise brand identities. This would still be fixable if we could put enough time into it. But looking at how little time we've put into this in the last 2 months, I could tell it was time to let go of this. We weren't ready for this.

We decided to stop Osmo. I'd love to explore this idea sometime in the future. But for now, we'll be focusing on different things. Osmo is something that will only work if both of us decide to put enough time into it and that won't happen for a while.

So what's next? I'm not really sure. I'm conflicted between starting a solo startup, cashing in on my skills and making some money, or trying to up-skill myself in some non-engineering skills. What I do know is, for the rest of 2025, I want to explore AI on a much deeper level.