Could Saline validators be replaced by ChatGPT?

The core idea behind Saline

In Saline, logic is installed on assets. Transactions can then be presented to the assets, and validator nodes determine whether each transaction passes the logic or “mandate” that gates the asset. If so, the validators vote to include the transaction.

‍

In practice, this involves writing an interpreter which is then run by each node to evaluate the logic. Could we replace this interpreter with an LLM? Let’s try.

‍

Architecture

Each node will now be a ChatGPT (or Llama, or whatever LLM you prefer) instance which will accept the transaction and its mandate, and determine whether the transaction should be accepted.

‍

The nodes then vote to accept or reject the transaction (for simplicity’s sake, let’s go with BFT’s 3n+1: 7 positive votes out of 9 nodes are needed to proceed with the transaction). 

‍

A simple example: limit order mandate

ChatGPT seems to be a fine interpreter for a simple price-based condition! Perhaps Cowswap should look at it for the evolution of the CoW Protocol.

A mandate with policy functionality: whitelisting and conditions

ChatGPT did very well in evaluating a policy mandate. Watch your backs, Fireblocks and Predicate.io!

‍

What about dynamic vesting?

We previously (https://www.saline.network/blog/dynamic-token-vesting-with-intents) discussed the idea of vesting investors as a function of how the chain is doing. Let’s see if ChatGPT can implement the vesting part of the mandate (for simplicity’s sake, we are not yet including the market making part): 

‍

Sablier (https://sablier.com/), take note…

‍

ChatGPT solver

Can ChatGPT be used to suggest transactions that everyone would be happy with?

‍

We need to find an optimal set of partial transactions where both sides agree on the price, then suggest an additional set of transactions to fully satisfy everyone's trade requests.

Step 1: Understanding the Offers

1. John wants to sell 1 BTC for ETH at ≥ 42 ETH/BTC.
2. Amanda wants to sell 28 ETH for SOL at ≤ 145 SOL/ETH.
3. Colton wants to sell 200 SOL for BTC at ≤ 670 SOL/BTC.

We need to match these trades while ensuring each party gets what they want.

Step 2: Finding the Best Partial Transactions

We notice that:

• Amanda has only 28 ETH, which is less than the 42 ETH John requires.
• Colton’s 200 SOL won’t generate enough BTC for John’s full trade, so we need another BTC seller.

Partial Trades That Work

1. John sells 0.67 BTC for 28 ETH at 42 ETH/BTC
   
   • He receives: 28 ETH
   • This is partial since he originally wanted to sell 1 BTC.
2. Amanda sells 28 ETH for SOL at 145 SOL/ETH
   
   • She receives: 28×145=4060 SOL
   • This completes her trade fully ✅.
3. Colton sells 200 SOL for BTC at 670 SOL/BTC
   
   • He receives: 200/670≈0.30 BTC
   • This partially satisfies John's BTC requirement.

At this stage:

• Amanda is fully satisfied ✅.
• John still needs to sell 0.33 BTC for ETH ❌.
• Colton still has 0.37 BTC left to acquire ❌.

Step 3: Additional Transactions to Complete the Loop

We need:

1. Another buyer for John’s remaining 0.33 BTC for ETH at 42 ETH/BTC
   
   • Potential solution: Find another ETH buyer willing to buy 0.33 BTC for 13.86 ETH.
2. Another SOL seller to provide more SOL to Colton in exchange for BTC
   
   • Potential solution: Find a new trader who is willing to sell around 247.9 SOL for 0.37 BTC at 670 SOL/BTC.

Final Answer:

Optimal Partial Transactions:

1. John sells 0.67 BTC for 28 ETH (42 ETH/BTC).
2. Amanda sells 28 ETH for 4060 SOL (145 SOL/ETH).
3. Colton sells 200 SOL for 0.30 BTC (670 SOL/BTC).

Additional Transactions to Complete the Trades:

1. Find a new buyer for 0.33 BTC at 42 ETH/BTC (needs 13.86 ETH).
2. Find a new SOL seller willing to trade ~247.9 SOL for 0.37 BTC at 670 SOL/BTC.

This fully satisfies all trade requests. ✅

‍