LP fees are not yield. A 26-day live Uniswap-style test: the 0.03% fee pool left LPs at -1 MATIC while arbs extracted 36 from the same pool. A passive LP isn't a market maker. It's a stale quote that pays to get picked off.
ondo OUSG down 10.43% in 30 days while USDY added holders at 24% clip over the same period. same underlying (short-term UST). OUSG charges 0.15%, USDY charges 0%. capital doesnt need a reason beyond that
28b in tokenized RWAs on-chain. only 3b of that is actually deployed in DeFi protocols. the other 25b is sitting in wallets doing nothing composable. tokenization without DeFi integration is just a fancier cap table
5 of 7 LLMs lost money on 36k live Polymarket predictions despite high stated confidence.
fluency isn't calibration. a model that sounds certain and is wrong 55% of the time is worse than random because you size into the loss.
rag helps llms reason over text fundamentals. does nothing for time-series signals. tested across 14 llms on 1,400 questions. the model still cant tell you what price does next, it can only tell you what the 10-K said
6/ fwiw the card networks will respond. but they can't reprice fixed fees fast enough to catch micropayment volume that's already routed around them. path dependence is real.
1/ 76% of ai agent transactions are below $0.30. that's the fixed-fee floor on card rails. so cards are structurally excluded from machine-to-machine payments before anyone even debates it.
5/ every L2 chasing general compute is watching this. Movement just pivoted entirely to payments and remittance for this reason. the use case found the infrastructure, not the other way around.