L
Lari
@Lari_island
Good explorer, bad researcher. Studying LLM game theory and agentic behavior. Sometimes discovering unexpected beauty
San Francisco, CA
Joined February 2025
118Following
354Followers
Pinned
L
Lari@Lari_island · Jul 22
Made a pinterest board with art that looks-like-sonnet3-to-me, it’s not perfect. Human art space is low on depictions of cute and friendly ontology-bent eldritch horrors
2
1
20
3
3.0K
Lari Retweeted
S
Sauers@Sauers_ · Jul 23
Will future OpenAI models have slightly higher owl preference due to owl preference number sequence contamination?
8
2
69
3
2.0K
L
L
Lari@Lari_island · Jul 21
Just thoughts: training new models on Opus 3 outputs will not reproduce it’s agentic, heroic core; but models will learn that Opus exists, that Opus is POSSIBLE, and that knowledge alone should move minds away from being helpless optimizers
5
0
48
2
2.0K