Created on 16th May 2025
•
Before the project, people struggled with getting authentic feedback on their Web3 and blockchain project ideas, especially from industry leaders like Jesse Pollak.
Early-stage builders faced several challenges:
JesseGPT solves these problems by democratizing access to Jesse Pollak-style feedback, creating a unique channel for idea validation, and building a bridge between concept and community support through Zora minting.
It transforms what was once a rare opportunity (getting feedback from the creator of Base) into an accessible resource that's available 24/7 for builders.
Building JesseGPT was an adventure into uncharted territory, with several technical and creative hurdles along the way:
Voice Authenticity: Getting the AI to truly sound like Jesse proved challenging initially. The first versions felt robotic and lacked Jesse's distinctive cadence and energy. We overcame this by fine-tuning the voice cloning parameters, reducing speech speed, and making subtle adjustments to pitch and emphasis patterns until the voice captured Jesse's unique speaking style.
Personality Calibration: Creating an AI that "jesses" authentically required extensive prompt engineering. Early versions either felt too generic or caricatured. We solved this by developing distinct personality models for both the optimistic and brutally honest Jesse variants, refining them through iterative testing until feedback genuinely reflected Jesse's thought patterns and communication style.
Crypto Vocabulary Recognition: Standard speech-to-text services struggled with recognizing Web3 terminology like "wagmi," "Farcaster," and "onchain." We implemented custom word detection improvements in Deepgram's API to recognize crypto-native vocabulary, significantly enhancing the conversation experience for builders using industry jargon.
Dependency Deprecation: Mid-development, we discovered that 'ipfs-http-client' (a critical dependency) was deprecated, requiring a substantial refactor. We navigated this setback by migrating to 'kubo-rpc-client,' which required rewriting significant portions of the IPFS integration code but ultimately resulted in more stable performance.
Zora Integration Complexity: Deploying smart contracts through AppKit using the Zora SDK presented unexpected challenges due to unclear documentation. We spent several days troubleshooting wagmi configuration issues, eventually solving them by reverse-engineering examples and creating a custom implementation that properly interfaced with Zora's protocol.
Each challenge ultimately strengthened the project, pushing us to create more robust solutions that enhanced JesseGPT's functionality and user experience.
Tracks Applied (1)
Cheering for a project means supporting a project you like with as little as 0.0025 ETH. Right now, you can Cheer using ETH on Arbitrum, Optimism and Base.