Inspiration

Software 1.0: Traditional code written by humans for computers Software 2.0: Neural network weights (parameters trained via optimization) Software 3.0: LLMs programmable through English prompts

Remarkably, we're now programming computers in English - the first time in computing history where programming happens in natural language.

-Andrej Karpathy

Natural language is now how we create. How we craft our written voice, the subtleties, the direction, the vision, and then we handoff to AI agents to execute the vision. The quality of our written language has become incredibly important to the quality of the output. Andrej himself noted that putting in the effort to create quality instructions increases the chances of success.

No more typing word by word or copying paste across the universe into a chat interface. We can turn language into lego blocks to rapidly speed up how and what we communicate to our favorite AI agents. The need for language composition tools is stronger than ever, and I believe this layer before AI execution will allow us to greatly improve generative results.

What it does

ConveyTo.ai is the professional IDE for AI communication - transforming chaotic prompt composition into a critical step in Human-AI workflows.

Stop losing your best prompts. Our three-panel workstation lets you craft, organize, and reuse AI prompts like a pro. Create dynamic prompts by referencing existing prompts with '@', and build a searchable library of your AI knowledge.

Key Features:

  • Smart Composition: Prompt references, and text editing
  • Professional Organization: Full-text search, tagging, and organization across your entire language portfolio
  • Universal Compatibility: Copy polished prompts to any AI tool - ChatGPT, Claude, Cursor, APIs

The result? Never start from scratch again. Transform from typing random prompts to building systematic AI workflows that compound your productivity.

It's the difference between coding in Notepad vs. Visual Studio - but for AI communication.

How we built it

I spent the first two weeks of the hackathon researching and reading to find something I really believed that I wanted to build. And then when the time came-- I basically lived in bolt.new for 12 hours straight. Started by describing my vision of "Visual Studio for prompts" and watched bolt.new generate the initial three-panel layout. To help direct, I'd sprinkle my favorite prompts like "Be a craftsman. Add subtle details, colors.."

My workflow:

  • Described features in plain English ("I want users to type @ and see a dropdown of their prompts") and add it prompts.
  • bolt.new would generate the code and integrate database.
  • I'd test it, find edge cases, then have a conversation about fixes
  • Rinse and repeat until MVP.

Challenges we ran into

The biggest challenge? It felt completely unnatural to be outside my IDE. I kept hovering over the export button, waiting for the moment I'd need to bail to VS Code for "real development"... but that moment never came. bolt.new simply handled everything I asked. There was a lesson to be learned of relinquishing my role as a developer and accepting a different role as I collaborated with Bolt.new.

I went from skeptical ("I'll need to handle auth myself") to amazed ("wait, it just built row-level security policies?") to totally converted. By the end, I wasn't thinking about exporting anymore - I was thinking in bolt.new.

Accomplishments that we're proud of

I'd say the vision. Often coding products and applications feels like a battle between UI/UX, Product development, Frontend and Backend.... but in this project, I spent a lot of my -- most of my time-- thinking about what I wanted to build, researching and reading... doing things I normally would not do at all if I was in a hackathon. This hackathon showed me a glimpse of what the future of human-ai collaboration might look like, where humans are building their visions and imagination, and AI is building them out.

What we learned

I learned how I want to incorporate Bolt.new into my workflows! Bolt.new is astonishingly powerful at quickly prototyping ideas for rapid MVP or brainstorming UI.

What's next for ConveyTo.ai

Please visit us at conveyto.ai to find out!!! We are very excited :) Partial Autonomy Slider This is one of the coolest yet simple projects you've seen in the hackathon. This one should win at least a shoutout. :)

Built With

Share this project:

Updates