The Journey of Alex Ruber and Parth Chopra
Alex Ruber, a 29-year-old cofounder and CEO based in San Francisco, has a story that highlights the challenges and triumphs of building a successful startup. Alongside his cofounder, Parth Chopra, they entered Y Combinator’s fall 2024 batch with an AI shopping tool. However, their journey was not straightforward.
After running at a loss, they pivoted through five to six ideas before finding success. Ruber shared that their biggest mistake during the search for a winning app was how they handled pivots. This as-told-to essay is based on a conversation with Ruber, offering insights into their experiences.
The Early Days and Challenges
About five months after being in Y Combinator, Ruber and Chopra thought they might have to give up. They met through Y Combinator’s Co-founder Matching Program, a free resource to help people find potential business partners, in January 2024. Both were working full-time when they started building projects together. They launched an AI-powered shopping tool and were accepted into Y Combinator’s fall 2024 batch.
It became clear that the shopping app was not economically viable and was running at a loss, with no clear path forward. They entered what Ruber calls “three-month pivot hell.” Within those three months, they tried about five to six different ideas and almost gave up. Then, they built a very simple game on a whim, and by August 2025, roughly five months after launch, their monthly revenue had reached $144,533.
The Struggle with the Initial Idea
Ruber and Chopra did not want to let go of their first app idea. Both were full-time engineers when they met and bonded over their love of building projects on the side. They are also both really into fashion, especially thrifting. They thought that the secondhand shopping market would only grow as concerns about the environment grew. Combining that with the advent of AI, they built a basic conversational search engine where users could type anything in natural language.
The engine would search resale sites like Poshmark, eBay, Mercari, and ThredUp, then compile all the results into a single page. They loved using the app and got some traction, but it was very expensive to run. People were browsing every day, but they weren’t necessarily buying. The commission they made on the final sale wasn’t great either.
Exploring New Ideas
They spent 2 to 3 weeks on each new idea. They cycled through five to six fleshed-out ideas, up to 10, depending on how you define an idea execution. They tried everything from a therapy product to a self-journaling product to ideas for dating apps. They even went into the B2B side of the insurance and paper industries.
They constrained themselves to build the solution in one to two days so that they could put it in front of real users through friends, family, Reddit, X, and online communities as quickly as possible. For each of these ideas, they went super hard for maybe two to three weeks. They did user interviews and determined the product’s success metrics, and they were ruthless about them.
The Low Point and Breakthrough
Right before they came up with their successful app, they hit a low point. They had cycled through so many ideas, and the metrics weren’t looking good. They were completely exhausted and didn’t know what to do. There was one night when Parth and Ruber decided to go for a walk. They both wanted to take a break and, honestly, get out of the startup world.
They took two or three days off to reflect on what went right and what went wrong for each pursuit. On the last day, they went back to the whiteboard. They decided they wanted to build in the consumer space. That’s one thing they learned from that whole pivot hell: they didn’t love B2B.
The Birth of Candle
They started making observations about the world around them. Being a founder is a very lonely experience, and they noticed that a lot of their friends and family were worried about them. Every time they checked in or called, they’d always be like, “Oh, we’re just working.” Same bland answer every time.
So they built a very early version of Candle, their social connection app, mainly to feel closer to their loved ones again. It gives a one-question-a-day prompt for both people to answer. They didn’t have any metric goals when they first created their best idea. They were so strict about the metrics during their constant pivot cycle, but they didn’t have metrics for Candle right away because it wasn’t one of the ideas they took seriously enough to define metrics for before they built it.
The Role of AI in Their Success
They published the app on the App Store in March 2025. They have a full-time team of four, and in their first year, they made over $1 million in revenue. They don’t throw AI at every step, but it does help them. They have a large question library and are actually opposed to AI-generated questions for their app. They feel like they come across as inauthentic, and this is an app for human connection, so they have human question writers.
However, they do use AI to edit grammar. They use Claude Code and Codex for much of their code generation and for building mini-games in the app. They don’t have a translator, so they’ve started using AI to translate into German, French, and Spanish. This has worked very well, especially using Gemini and Claude. They also have AI recommendations for activities that two users could do together.
Lessons Learned
Using AI to recommend content or information that helps two people connect is how Ruber thinks AI should be used, not to replace the connections humans form with each other. They pinpointed their biggest mistake and learned from it. They feel so passionate about their app. Looking back, the biggest mistake they made was making an industry pivot that was too large each time they explored a new product idea.
Instead of picking a problem space they really love, trying their solution, and, if it didn’t work, switching to a different solution, they kept switching industries. They lost their passion for what they were building as they did that. Ruber’s advice would be to pick the problem space and not change it very easily. How you approach that problem is what you should be changing.
