Prompt3

Prompt3 uses Decentralized Prompt Engineering to create Specialized AI Assistants that give exceptionally good answers for specific subjects. The SuperPrompts that drive these Assistants come from Prompt Engineer submissions, who are paid, and have the ability to verify when their prompts are used in training.

The Customer experience is simple: choose a Specialized AI Assistant (ex. Paris Guide, Solidity Developer, Pregnancy Nurse), and pay some tokens to ask a Question and get a SuperAnswer. It’s way more effective than using ChatGPT by itself, and we want to have lots of these Specialized AI Assistants.

The Owner of the Specialized AI Assistant receives revenue from those Questions. The Owner is also responsible for choosing which Prompts to use in the generation of a SuperPrompt. With an array of Prompts selected, we used ChatGPT to go and create a SuperPrompt using those inputs. The more submissions that come in, the better the SuperPrompt should become. The Owner pays the Prompt Engineers when their Submissions are used.

Prompt Engineers Submit prompts, receive revenue when those Submissions are used, and have the ability to Verify that their Submissions were used. They have a page where they can check whether a Submission is included in the Merkle Tree of Submissions that were used in generating the SuperPropmt. The Merkle Root Hash, as well as hashes representing the SuperPrompt creation are anchored into a Smart Contract on Gnosis Chain by the Weavechain node that runs the computation, to guarantee =immutability and fairness.

The frontend of the application is a simple React app, with plenty of JavaScript code, and Web3.js for handling things like MetaMask login.

Weavechain is the core technology that handled data flows, secure token payments between users, the creation of the computational lineage, NFT/POAP access gating, and anchoring hashes to Gnosis Chain. We packaged the creation of SuperPrompts for example into docker images that used bash scripts and python code. This way Weavechain was able to run the docker image, and take a hash of it, to keep as a computational lineage piece.

Obviously we used ChatGPT’s API for both the SuperPrompt engineering, as well as for the Customer Questions and responses.

Finally, we used Gnosis Chain so that we could anchor the hashes into a smart contract on a public and permissionless network. We also leveraged POAPs as NFTs the be used for access gating, so that we could enable special functionality for the AI Assistant Owner role like generating SuperPrompts.