• 

  VoiceScope

• 

  Novus Dashboard

• 

  Novas Dashboard Memory

VoiceScope

As a solo founder, you have to juggle many hats all at once while also talking to customers, understanding their pain points, and iterating on your product based on their feedback. That means fitting time slots into an already hectic schedule to talk directly with them over calls. Scheduling is its own challenge since you have to make sure both parties are available at the same time. During the interview, unless you have an AI note-taking app, you have to take notes yourself and then summarize them into tasks or tickets that help refine your product. This adds pressure and slows down your build.

VoiceScope lets you focus on building and shipping while an agent runs the interviews and collects feedback for you, sending findings straight to the tools you're already using.

VoiceScope isn't only for builders either. Researchers who want to learn how people use a product can assign participants and let it run async interviews too.

Inspiration

In early 2025, I spent six months conceptualizing and building an AI chatbot for meal planning. The entire time, I was solely focused on building what I thought would fit my ideal users. I was proven wrong after I shipped it. The format was wrong, the experience was wrong. I learned that building in public and iterating on real user feedback was the right way to build but finding users and collecting feedback was a lot for one person to handle alone. That's how VoiceScope came to be.

How I Built It

I built VoiceScope with ElevenLabs powering the voice agent that runs the interviews. When an interview ends, the agent automatically generates a research brief and saves it to Notion, and files pain points as issues directly in Linear. The whole thing is designed to run without any involvement from the researcher once the link is shared.

Challenges

The hardest part was working with a voice agent for the first time and making sure the agent calls tools at the right time. Since the agent calls tools after the conversation wraps up, I kept running into trouble with the timing — tools would get abandoned mid-execution and the session would stall. On top of that, I wanted to make sure the app was actually usable end to end, so connecting the agent to Notion and Linear was another added challenge.

Accomplishments

The moment that surprised me the most was watching the agent extract real pain points from an actual conversation. I ran an interview from the perspective of a user of the meal planning chatbot I had built, and the pain points the agent surfaced were exactly what real users had told me before.

What's Next

VoiceScope is still early and there is a lot more to build.

Before the interview even starts, I want it to do pre-research on the product through web search or uploaded documents, so it goes in with prepared questions and target points rather than just a goal.

I also want to cnnect more tools beyond Notion and Linear so findings can land wherever the team already works.

Right now, sharing the interview link is a manual step. If VoiceScope could pull from an existing user list and send invites automatically, the whole process becomes truly hands-off, from interview to findings, with no manual work in between.

Built With

• clerk
• elevenlabs
• linear
• next.js
• notion
• redis
• typescript