@mmdhrumil@annanay Latency races usually don’t contain a ton of information, if everybody is racing to cancel / pick off the price was already discovered to some degree.
BREAKING:
@Coinbase has integrated Jupiter's Metis API to power its Solana token swaps.
It joins Uniswap, Robinhood, and dozens of others who trust Jupiter to deliver a world-class swap execution for their users.
Another company who knows to Just Use Jupiter.
The instant one aggregator segregates well, all others will quickly be forced to or start losing flow - market makers will provide MUCH tighter to known soft flow and widen on or drop aggregators that don’t do the filtering.
Somebody semi replicated prop AMMs on openbook, and despite having the lowest CU oracle update by far had to turn off as takers got better.
Next step in the game is going to be segmentation by the aggregator
Solving toxic flow using order flow filtering and segregation: (Idea #3)
This is the real benefit of Prop AMMs and the number one reason they are successful.
The biggest issue, by far, for any market-making operation is toxic flow or as some like to call it "adverse selection". That is, getting trades from other bots/professional trades that have an "advantage". I don't want to give an exact definition to "advantage" because it could be anything that lets them know the future price before you. Future price could be literally future price or they were able to trade against you before you were able to update your price, which is effectively the same except it's not about predicting the future relative to real world time but relative to the time from your price perspective.
I talked about this in detail in the last talk, but I want to mention just one point again from that day: There is a good reason 22% of the revenue of Robin Hood comes not from Citadel paying them for their retail non-toxic order flow (Payment For Order Flow PFOF).
With a fully custom execution, you can do a lot to try to filter out toxic flow and find the "good" order flow. In case that wasn't clear, I'm saying they give out different prices based on the incoming order flow type.
This is a long post about the rise of "prop AMMs", the history behind them and their advantages/disadvantages. This post will be different than my usual style of trying to be concise and is more of a story since it was meant to be for a DeFi Day talk like I did in the past 2 years. Unfortunately, this year I was asked to provide my ID/KYC (not the fault of the organisers) to be able to speak which I'd rather not do.
As a gift to the Solana community for Breakpoint 2025, we've decided to open-source `peephole` https://t.co/NVGSJ9ve62
peephole: Zero-copy account parsing for Solana programs
solana_program::entrypoint::deserialize allocates `AccountInfo` structs and copies account data. peephole skips that with a zero-alloc and zero-copy view over the account and instruction data.
Real-world: 51 CU oracle update, 67 CUs for 3 oracle updates, 12 CU trading contract, and many others.
h/t @zft_send
@liquiditygoblin Something something “the bureaucracy grows to feed bureaucracy” but these roles also have an important function as an accountability shield. “Please don’t ask the dev team about why Q is still broken. Message me so I can put it in the jira list for planning poker next week”