HomeLLM

1) A new way for LLMs to visually express themselves to better support early years autism. 2) A new social experience within Llms and agents 3) Created a concept to make ads feel native within LLm.

Comment 1

Inspiration

LLM chatboxes are incredible tools for gaining insights, exploring topics, and creating content. However, a significant disconnect remains between those who use them naturally and those who struggle to start. To bring the next billion users into the world of AI, I wanted to create the next step from a product perspective, designing for better accessibility, experience, and human connection.

The First Challenge: Giving AI Body Language

The first step was giving LLMs a new form of expression. While text and speech advances have been remarkable, I asked: What about body language? As visual creatures, this missing piece creates a fundamental disconnect. Current platforms recognise this and use emojis as a stopgap solution. My goal was to move beyond this.

The Trigger: Accessibility and Empathy

The initial trigger and foundation for this project was to move beyond expensive robots like QTrobot to help young people with autism develop social skills through consistent, accessible, and affordable interactions with an LLM chatbot. This led to a deeper thought: when an LLM uses visuals to react, how does it reason between showing visuals that are empathetic, providing a distraction, or progress the conversation.

The "Visual Erasing" Effect: A New Native Ad Model

The key question became: how to implement visual expressions beyond emojis in a natural way. My design inspiration comes from simple observations of human behaviour. I recalled watching those extreme window cleaners who climb buildings. The soap is applied and completely blocks the window glass, then is slowly wiped away, or erased, before revealing a clear view. This inspired the "visual erasing" effect you see. Visuals, whether emotional reactions, educational explanations, or ads, appear first, then gracefully make way for the live-rendering text answers. This innovative approach is only justifiable within LLMs due to their unique real-time text rendering, making it a truly native experience.

(To see these emotional reactions, use trigger words like "I know it's late," "Do bees drink," "Not good enough," or "Your answers are boring" and click the 'Think Mode' button.) I know it's late question, create time-sensitive visuals, for example, 3 am questions.

Solving the Ad Problem for LLMs and Developer Tools LLMs will eventually have to implement advertising, but the traditional route would be devastating to user trust. This new, native form creates a symbiotic hybrid between visual content and text answers, bridging the gap between traditional 'blue links' and the future of AI.

This model is especially powerful for "vibe coding" on AI tools like Bolt.new. It introduces "Code-Contextual Dissolving Ads", a monetisation layer that respects a developer's flow state by transforming ads into helpful, native suggestions: -Empathetic Interventions: A "Take a break" ad from a beverage partner becomes a wellness nudge, not a distraction. -Intelligent API Suggestions: When the AI generates code requiring a backend, a Supabase ad appears as a solution. This model allows platforms to monetise their free tier through hyper-relevant, value-added partnerships, solving the freemium dilemma without alienating developers.

What Else HomeLLM Does

A Creator Economy: Artists design and tag animations with emotional metadata. The demo uses Gemini API and Supabase to show and download an Emotion.json file. Try it out. Then, using a blockchain like Algorand could verify art ownership for greater creator protection. The visuals I created are okay compared to what artists could really achieve. All visuals were created before Midjourney shipped Midjourney Video; therefore, what can be really created is now limitless.

A Visual Chat Library: Previous chats are stored as visual "book covers" with motion and ElevenLabs speech, enabling easier accessibility, long-term connection, and sharing. You can share your visual "Home" with others, who will need permission to view a book, a significantly better experience than sharing a list.

Proactive & Personalised Homes: The Home environment is personalised. For free users, the Home may change to reflect non-intrusive sponsors. Hovering over the Home and Shop buildings triggers a tailored ElevenLabs conversation, inspired by dynamic billboards.

Serendipitous Social Connections: The platform facilitates human connections based on the questions asked. Users can interact with others in chatrooms dealing with similar issues through a tailored AI matching system. No invites, no manual joins, just connection through curious questions.

Your Table

Almost all items are clickable to a site, looking for a Bolt Cap? The purpose was to create a space for all your tools.

How I Built It

As a non-technical solo builder, I spent a significant amount of time on research and ideation. Allowing time to think things through was critical before execution.

Challenges I Ran Into

My initial goal was to start a team for this marathon project. When that didn't materialise, I decided to go it alone. The biggest hurdle was balancing my ambitious vision with the technical constraints of being a solo builder using a new tool. I have heard horror stories about how people have lost all their work. This was my biggest fear, I had little experience with coding. However, Bolt.new was instrumental in helping me achieve my design goals safely.

Accomplishments I'm Proud Of

I'm most proud of creating the 1% preview and how it adds to the aesthetics of the Ui. Also, "Treebond." This feature uses six emojis, from a seedling to an Oak tree, to represent the bond a user develops with their AI agent through deep conversations and frequent, helpful interactions. This visual feedback loop is the missing link that can help users trust the agent they're speaking to. The Treebond emoji will even devolve if engagement stops, mirroring real relationships.

What I Learned

This project was a powerful reminder that good design comes from learning from human behaviour. Tools like Bolt.new can empower individuals to bring complex visions to life with minimal compromise. Most importantly, I believe emotional intelligence is AI's next great frontier, especially in helping those with invisible disabilities.

Final Words

The overall goal of this project was to try and do something that feels different. For years within my Medium articles, I just enjoy coming up with new concepts that can make a difference for humans. Mostly ignored, but it's part of my DNA. So this Bolt.new Hackathon was a stage to push limits.

Built With

Share this project:

Updates