It has been an extremely active 24 hours for FrankenTUI development, with countless new features and functionality added (including massive numbers of tests and verifications). Have you ever seen a terminal like this? Because I know I haven't! Can't wait to start building stuff!
@annanay RFQ in crypto originated for long tail assets without CLOBs as a mechanism to bootstrap liquidity for the infinite design space of tokens. tardfi largely has a bounded set of tickers (excluding options, SPVs)
there's more than 1 use case for most things, including RFQ
i can admit when i was wrong.
so there's good news and bad news:
bad news: sniping modern pateks/rolexes on ebay is not a thing. it's an ~efficient market
good news: passing unidetified manhattan estate liquidation lot images through a vision model to find mispricings appears to be a thing
the core of the thesis has evolved to: find auction pools where the clearing price is structurally depressed for reasons unrelated to what the item is, and value it against the current bid
you thought C-tier perp DEX RWA flow was soft?
how about regional pickup-only estate liquidation listings
that busted up old japanese paper divider that sat next to Gramps' writing corner? it's worth $12,000 to a collector, but the top bid is for $30, and the seller just wants it gone
Maw Maw's favorite brooch listed for $120? it's tiffany and it's worth $25,000
but nobody knows these valuations except a few thousand people in the entire USA per niche category. and the odds that they are going to sift thourgh hundreds of thousands of random listings daily that are mislabeled (or, more commonly, unlabeled) to find all mispricings is ~0
so i (claude) built a tool to scrape nyc tri-state area estate liquidation sales, price listings, and surface the gems
first pass does a claude haiku 4.5 valuation. 2nd pass (manual) is an opus 4.7 valuation for the high ticket items with low bids. on many thousands of listings per day
manhattan is a particularly good sub-market for this (although probably more efficient than some) because of the concentration of wealth
but any major metro will have some inefficiency / insane deals
--
to expand on the thesis a bit
edge comes primarily from the venue, not the query. we are trying to stack multiple qualifiers from the following list to determine likely 'soft' venues for listing discovery
A. limited bidder pool - small venue, obscure platform, in-person, local-only
B. wrong bidder pool - charity gala, corporate event, room full of non-specialists
C. friction - pickup only, wire/cash only, old bad website
D. non-monetary motivation - tax deduction > price, speed > price (estate, divorce, bankruptcy)
E. information asymmetry - seller doesn't know what they have
F. time-boxed release - court deadlines, closeouts, lapsed reservations
scraping currently on estate liquidations in the tri-state area but i think there is probably softer flow out there. maybe charity auctions or court-ordered liquidations
----
good side project, will continue
@cryptopunk7213 now i think about it, culture has parallels to an llm. its basically just the consensus behaviour in a society. societal norms the same - the accepted default behaviour. a normie = consensus being (determistic)
@cryptopunk7213 what if, culturally, we behave non deterministically because we live in a people-driven world. but moving into a post-AI world, we start to behave deterministically. you already see this somewhat since the internet - culture transcends borders and we're all kind of blending.
@doodlestein a huge whitepill for everyone else is that a multi-billion dollar company is making schoolboy errors like this. you're probably overthinking everything, get to work.
@robj3d3 My family is cypriot, have some great memories holidaying there when younger. The best european food there is as well in my opinion (italy close second)
@BeyerSebastian@kr0der how did you infer this from the OP? you don't write tests to test the amount of prisma queries you're doing. the result is correct with terrible performance
Great skill if you're working on multiple worktrees/projects. Reserves 10 consecutive ports so there's no conflicts and agents don't end up killing eachothers processes.
by @turingou