Founder · 2023 – 2025
Building an AI assistant for ADHD, solo
Summary: Through the Hook and Antler incubators, I founded a startup building an AI personal assistant for adults with ADHD, by text and voice. I did the research, positioning and go-to-market plan, and built the MVP myself with AI coding tools.
The problem
People with ADHD often know what they need to do. What breaks is remembering it at the right moment and following through. Most productivity apps add more lists to manage, which makes it worse.
Key product decisions
- Telegram as the interface. No new app to install or remember to open. The assistant lives where users already chat.
- Why Telegram over WhatsApp. Telegram has a smaller user base, but building a bot is easy and free, which matters most for a rapid MVP that has to validate ideas.
- Voice as well as text. Saying a thought out loud has less friction than typing it, which matters when the thought is about to slip away.
- The assistant reaches out first. Reminders and check-ins come to the user, because waiting for the user to open the tool is the exact failure mode.
How I built it
| Layer | Tools |
|---|---|
| Interface | Telegram bot |
| Intelligence | OpenAI and Anthropic LLM APIs |
| Voice | VAPI |
| Building | Cursor, Claude Code |
30-second demo video coming soon
What I learned
- Testing with pilot users taught me more than any planning doc. For example, I learned to favor voice over text: users were reluctant to type replies, but much more willing to pick up a call from the assistant.
- Building it myself changed how I write specs: I now know what's quick, what's hard and where LLMs fail.
- Getting real retention is a much harder problem than building a working product.