Launching Hovrlay

For a long period of time, I kept watching people do strange things to get through interviews. AI bots slipping into Zoom calls under names nobody recognized. Phones propped against laptops at angles that fooled no one. Friends on standby in a Discord server, listening in to the call and answering questions discreetly. To me, the problem stood with how inefficient the way people sought help was. The help wasn't the issue. The way they were sneaking it in was.

The Setup

At the time, I was a SWE at a startup that had millions of monthly active users. I was good, the pay was good and the trajectory looked respectable. But the longer I stayed, the more I realized that corporate success has less to do with building great things and more to do with justifying your existence to the people above you. Visibility matters more than depth. People aren't building because they love the craft anymore. They are optimizing for performance cycles, promotions and survivable careers.

So on the side, I had formed “Goo Goo Garbs”, a baby clothing brand in the US. I had started it with real hope, taking care of sourcing, shipping, company incorporation, marketing, payments and everything else. After over a year, it was not working and I knew it had to be shut down. The logical next move felt obvious — get a higher paying job, reset and build something again eventually.

I started interviewing at Google.

I didn't grind for it. I want to be clear about that part. I had already spent years deep in competitive programming, data structures, algorithms and system design. The interviews felt natural even, and within three months I cleared seven rounds of interviews making the cut to team matching. It was here where one conversation turned design heavy, something team matching rounds aren't supposed to be about . I walked out unsure so I waited.

A few days turned into weeks and then one morning, the rejection call came.

A year of Goo Goo Garbs did not work. Google did not work either. And I was tired. Not because of the struggles but because I knew I was good and yet my capability did not matter. Somewhere under all of it, a thought started to take shape in my head — “If only I had a cleaner way to get help in my interviews.”

The Spark

In January 2025, Interview Coder went viral. Roy was doing it all openly. Walking into Meta, Google and Amazon with the overlay visible on his screen, posting about it, almost daring anyone to call it cheating. And the strange part was that the internet wasn't condemning him, quite the opposite, it was cheering him on.

There was an irony there I couldn't quite unspool. I think I admired him for it, even while it stung. He'd just made visible the exact thing that, if I'd let myself reach for it a few months earlier, might have changed the way that rejection email landed in my inbox.

The tool itself bothered me. It was elegant. No bot in the call, no invite link, no guest at the door, just a window on your screen that only you could see and is invisible to screen share. It worked using screenshots of whatever is on your screen. Static. A frozen frame of a moving conversation. What if the interviewer speaks the problem statement rather than showing you a question on the screen? I kept thinking what if it could actually listen? What if it understood not just the question on the screen, but the whole shape of what was being said, as it was being said?

Same Idea, Different Zip Code

Eventually Roy got expelled from Columbia and went through a phase of his own. After a few months, on 20th April 2025 he launched Cluely. Roy raised $5.3M in funding and assembled a team. I read the announcement on my phone and felt two things at once. The first was a flicker of validation, I wasn't crazy, I'd seen the same thing he saw. The second was more serious. The problem I had convinced myself was worth solving was now being attacked by a company with venture funding, elite talent, virality and momentum on its side. Meanwhile I had a laptop, a Cursor subscription and a half finished overlay running on my desk.

There's a specific kind of weight that comes not from self doubt but from arithmetic and that weight lived on me for a bit. A funded team moves faster, ships more, markets louder. I started building anyway. Not because I had a plan for how to win but because stopping felt worse.

The Grind, Then the Quiet

I was building Hovrlay on the side while still at my job. My then manager, who had a way of making you feel like your best work would never quite be enough, made the decision to quit my job easy for me. I put in my notice, packed up Bangalore and moved back home to Mumbai. No rent, home food and my own space to build, definitely a cheat code.

If you're in your early twenties and you have the option to work from your parents' place for a while, take it. Hovrlay had my full attention.

For the next stretch, I stopped thinking about outcomes. Money stopped mattering, launching stopped mattering. I just fed my brain whatever it wanted next. One task, then the next, then the next. Building the Electron app, setting up the server infra, setting up CI/CD pipelines, transcription research, UI design, the website, GTM strategy, Apple approvals, Google verification, onboarding flows, monetization flows. All of it, alone, in no particular order, with no particular plan. Hours and days disappeared. There's something almost holy and dangerous about that state. I have a lot of affection for the version of me that lived inside those months but I wouldn't recommend being him.

One day though out of these many days, I stopped. No story there, no crisis point or clarifications. But the laptop remained shut and so for 2 months. I was already hitting the gym but over that period of time I picked up MMA, deleted Linkedin and X and stopped caring. Burnout, when it actually arrives, is not very cinematic, but rather messy and full of belittling questions that shut you down and in an attempt to hold onto any sense of worth in myself, I gained ten pounds of muscle. If I were to put it plainly, one day I wanted a million users, the next day I wanted bigger arms.

Unfinished Business

The thing about those two months away from Hovrlay was that I never really left it. Yes, I stopped coding or opening the repo. Stopped telling myself I was working on it. But the product stayed at the back of my head. Every time I came across another AIinterview tool, my brain would drift into implementation mode. I’d start thinking about latency, edge cases, conversation flow and tiny UX decisions that anybody else looking at it would barely take notice of.

So the laptop opened again. At first it was supposed to be temporary. Just clean up a few rough edges. Fix a couple things that had been bothering me. But somewhere during those late night rebuilds, the product changed. Or maybe my understanding of it did. Earlier versions of Hovrlay felt like prototypes trying to justify their existence. Every feature felt eager. Overexplained. Slightly insecure. This version felt different. The product became quieter. Faster. More deliberate. I started deleting things aggressively. Simplifying flows. Tightening the overlay. Reducing friction everywhere I could. For the first time, the product stopped feeling like an experiment and started feeling inevitable.

The Launch

Hovrlay has officially launched on macOS and Windows and you can try it for free. No bot joins your interview call, no phone beside your laptop, no friend feeding you answers from another microphone. It will sit on your screen, listen into the conversation, understand the context all in real time to give you the help you need. That's the entire idea.

In a way, this project started long before I wrote any code for it. It started the moment I realized people weren't cheating because they were lazy. They were doing it because interviews had already become a performance of optimization and everyone was solving for survival in their own way. Hovrlay was my answer to that feeling.

If you're building something alone, somewhere in a room nobody will ever see, I hope you keep going.