Devin Conley pfp

Devin Conley

@dcon.eth

236 Following
1405 Followers
Devin Conley pfp
Devin Conley
@dcon.eth
The upgraded P4 homepage now shows 📈 trending market data ⚖️ your recent custom comparisons 📜 top market reports Try out the live beta today.
0 reply
0 recast
4 reactions
Devin Conley pfp
Devin Conley
@dcon.eth
@tempetechie.eth @matthewfox sorry I think you guys got scammed
1 reply
0 recast
0 reaction
Devin Conley pfp
Devin Conley
@dcon.eth
just got back from europe, now cranking the air conditioning and flushing entire rolls of toilet paper with our industrial strength plumbing to celebrate the 4th of July
3 replies
2 recasts
17 reactions
Devin Conley pfp
Devin Conley
@dcon.eth
are weights open source? could build some kind of optimistic consensus to let users verify results and dispute if needed. or zkml proofs, but that's expensive
1 reply
0 recast
1 reaction
Devin Conley pfp
Devin Conley
@dcon.eth
agreed there's a lot of room for cool work here overall, I think that topic modeling/clustering is a relatively easy ML problem (for short or long form) but quality scoring is still really challenging/nuanced, because it so heavily depends on evolving taste and trends there's probably an interesting hybrid approach to predict quality with a mix of social graph data + post content
0 reply
1 recast
0 reaction
Devin Conley pfp
Devin Conley
@dcon.eth
definitely depends on the audience, but a raw demo implies no wasted energy on frills/handwaving
0 reply
1 recast
0 reaction
Devin Conley pfp
Devin Conley
@dcon.eth
definitely depends on the audience, but a raw demo implies no wasted energy on frills/handwaving
0 reply
1 recast
0 reaction
Devin Conley pfp
Devin Conley
@dcon.eth
I like live demos where it's just a dev screen sharing their terminal
1 reply
0 recast
4 reactions
Devin Conley pfp
Devin Conley
@dcon.eth
sir please be careful
0 reply
0 recast
0 reaction
Devin Conley pfp
Devin Conley
@dcon.eth
agreed there's a lot of room for cool work here overall, I think that topic modeling/clustering is a relatively easy ML problem (for short or long form) but quality scoring is still really challenging/nuanced, because it so heavily depends on evolving taste and trends there's probably an interesting hybrid approach to predict quality with a mix of social graph data + post content
0 reply
1 recast
0 reaction
Devin Conley pfp
Devin Conley
@dcon.eth
LLM prompting
0 reply
0 recast
1 reaction
Devin Conley pfp
Devin Conley
@dcon.eth
... and the best library makes you a helpless developer
0 reply
0 recast
0 reaction
Devin Conley pfp
Devin Conley
@dcon.eth
I just rewrote all of the sql logic to build this table and the issue was actually just a lower case string
0 reply
0 recast
2 reactions
Devin Conley pfp
Devin Conley
@dcon.eth
I think sub ~10 seconds is my threshold to maintain some kind of flow (programming or otherwise)
0 reply
0 recast
0 reaction
Devin Conley pfp
Devin Conley
@dcon.eth
conditioned on the degen tipping meta
0 reply
0 recast
4 reactions
Devin Conley pfp
Devin Conley
@dcon.eth
definitely helps! (until we inevitably hit some subtle silent bug from a difference between the cloud vs local versions of the thing)
0 reply
0 recast
0 reaction
Devin Conley pfp
Devin Conley
@dcon.eth
cool visualization, does this exclude smart contract accounts? wonder if these patterns could be related to some defi protocol or programmatic strategy
0 reply
0 recast
1 reaction
Devin Conley pfp
Devin Conley
@dcon.eth
(I end up distracted on farcaster complaining about it)
1 reply
0 recast
2 reactions
Devin Conley pfp
Devin Conley
@dcon.eth
dev flow with a long feedback loop is brutal... infra deployment, CI systems, data pipelines, etc
1 reply
0 recast
8 reactions
Devin Conley pfp
Devin Conley
@dcon.eth
my love language is filenames with no spaces/special characters
0 reply
0 recast
4 reactions