Content
@
https://opensea.io/collection/dev-21
0 reply
0 recast
2 reactions
Chinmay ๐น๏ธ๐ฟ
@chinmay.eth
Some technical lessons we learned while building @basketdotfarm Thread Below ๐
2 replies
2 recasts
12 reactions
Chinmay ๐น๏ธ๐ฟ
@chinmay.eth
1. AI is easy to get started with and hard to get accurate. Using any existing frameworks (we are using Eliza), you can create a demo project within hours. But when you try using it with actions and accurate user flows, it's very hard to perfect.
2 replies
0 recast
7 reactions
Chinmay ๐น๏ธ๐ฟ
@chinmay.eth
2. Too many tokens and not a single reliable source of truth We have millions of tokens on @base. Many have the same ticker. It's very hard to guess which one is user looking for. Our current solution - reverse sort by the amount of transactions - pick the top one and ignore all else - hardcode this so that AI does not screw up
2 replies
0 recast
4 reactions
Chinmay ๐น๏ธ๐ฟ
@chinmay.eth
3. Gas estimations are still a problem. After spending 8 years in crypto, I was hoping to rely on something like Alchemy to be accurate with gas estimations, but I was wrong. Sometimes gas estimations are off by 10x. This can scare the users and cancel the transaction altogether. Our solution: - manually enter gas estimations. This will definitely backfire in the future, but it will work for 90% of the transactions.
1 reply
0 recast
5 reactions
bry
@brykayne
was wondering if volume / holder count would be helpful here too definitely hard coding the top 100-250 feels like it makes sense
0 reply
0 recast
0 reaction