How Riel Chatbot Grew — Accessibility Born from a User's Request

"The text is too small, I can't see it well." One day, I received a request like this. It was a feedback from an elderly family member who used Riel daily. That day, I immediately built a text size adjustment feature. Many of Riel's features were born this way — not from my roadmap, but from someone's **"This is inconvenient."** This post is a bit different from the previous development logs. **Putting aside the detailed technical know-how, I'll focus on the broader strokes of growth** — how requests turned into features, and in which direction Riel is growing. > This is the story of **[Riel](/chat)**, an AI chatbot built by a solo developer in Korea. All user personal information has been anonymized. ## 1. Growth Came from 'Requests,' Not Plans It wasn't a case of drawing up a roadmap and following it in order. **If someone said, "This is inconvenient," that became the next thing to build.** So, Riel's growth is a **virtuous loop** — ## 2. Real "Requests → Solutions" — Directly Implemented Below are all **anonymized actual requests** I received (personal information excluded). - **🔍 "I'm old and the text is small when I search"** → **3 levels of text size.** Pressing '가' in the menu increases it by one level, and it stays that way once set. - **🔊 "If I ask something in text and my eyes are tired, can you read it out loud?"** → **Read answer aloud.** With one button, Riel reads the response audibly. - **📺 "It would be great if I could refer to YouTube videos too."** → **Include relevant YouTube links in answers.** It finds helpful videos to share if available. - **🔎 "As conversations pile up, I can't find what we discussed before."** → **Search conversation history.** You can find past discussions not just by title, but also by keywords within the conversation. - **🧭 "I consulted Riel about a relative's or friend's problem, and Riel remembered it as 'my problem'."** → **Distinguish between self and others.** Now, it understands others' issues as not its own. - **🖼️ "I could only upload one photo, videos weren't supported, and it was hard to describe things in words..."** → **Upload multiple photos at once + analyze album videos + ask by drawing a picture.** - **🟢 "Older people are more familiar with Naver than Google."** → **Naver Login.** These might seem like small requests, but they've accumulated to make Riel a chatbot that **"even genuinely elderly people use daily,"** not just a developer's creation. ## 3. The Big Picture — Riel's Four Growth Directions Setting aside the detailed implementations, the **main pillars** of Riel's current growth are: - **🌱 Remembers** — The more you use it, the more it gets to know you. Unlike chatbots that treat every interaction as new, Riel understands you better as conversations build up. (Riel's key differentiator) - **🎙️ Listens and Speaks** — Converse with voice, and get answers read aloud. This makes it usable even for those with hand or visual impairments. - **🖼️ Sees** — Understands photos, videos, and sketches. You can show it things that are hard to describe with words. - **♿ For Everyone** — Text size, voice, familiar login methods... regardless of age or device. These four directions are shaping Riel into **"an AI by your side,"** rather than just a "smart AI." ## 4. So — The True Engine of Growth Looking back, Riel's growth engine has been **the users, not the models or my ideas.** A simple "Thank you for always implementing my requests," or honest feedback like "This is inconvenient" — these were the seeds for the next features. > **Listening to a user's small inconvenience is more important than one well-made feature.** Riel will continue to grow following the requests of those who use it. --- Thank you for reading this far into the development log series — from blockchain, Polygon purchases, AI video generation, and Riel's growth. All of these were **real records of hands-on experience.** If you're curious about Riel, **try it for free](/try)**. It's still growing. 🌳