I the analogy that GPT is improvising/speaking, like you would in a normal conversation. When I talk out loud (quickly), I only have ~1 word of lookahead just like GPT. But if I want good answers to hard questions, I need to slow down and actually write[1] something down. So the wrapper script that iterates/recursive on responses that OP describes is analogous to the slowing down process of writing.
I'm also curious what sort of results the iterative process can lead to. The movie script example in OP is impressive, but does it reach a stable state? Does it work for other types of prompts (coding related/other) that i've seen on twitter? All very interesting.
Monaco. Top down heist simulator that’s very silly with amazing music.
Minecraft with buddies online is amazing.
Towerfall ascension campaign. Made by developers of Celeste (so it has A++ feeling controls). Single screen single hit kill defend against waves of enemies in several worlds. Good difficulty progression. And 4 player VS is a blast.
Not coop but vs:
Duck game. Works best with 4 ppl, high skill based, super silly single screen instant hill tournament. Playing this with buddies is consistently the most fun I have had in 10 years of playing games.
Gang beasts. Very very silly, not skill based last man standing brawler on interactive maps.
Party games like jackbox can be super fun, including gartic phone, skribbl.io, cards against humanity.
This is very interesting, thanks! It reminds me of a friend's adventure, he found a contract containing some ETH that could be called with some more ETH and would send the entire sum back to the caller. He analyzed it a bunch of ways to make sure it wasn't a scam, and then sent some ETH to it.
It was a scam.
I wonder how it was done, Etherscan didn't show anything and compiling it led to a few bytes of difference between what was compiled and what was deployed.
There are so many clever ways to code honeypots using obscure peculiarities of Solidity and/or Etherscan that there's little hope of being sure that it isn't a scam just by looking at the code and transaction history.
Fortunately, there are tools like Ganache, which you can run with `ganache-cli --fork` to reliably emulate locally what will happen when transactions are sent to mainnet. I would accept no substitute approach when dealing with suspect contracts.
Interesting, it might use a flaw in the Etherescan contract verification[1]. But in any case, when you expect a honeypot you can and should execute the contract off-chain[2] and examine the resulting state (specifically your account balances) before committing a real transaction. Wallets should really do this by default, but unfortunately there doesn't seem to be a lot of resources available for common goods projects like wallets, so we are stuck with primitive tools.
