"Did I formulate a hypothesis before asking? Before I run an agent on any non-trivial task, I write down, even just in my head, what I think the answer might be. It's OK if it's a bad guess. You're always welcome to change your mind in light of new information. Once you've read a well-written response, though, it's hard to imagine any other possibility. If you have no guidance whatsoever, ask for what Thariq Shihipar calls a blind spot pass: give the model what you know and have it find your unknown unknowns."
The time everyone start to do this...
The time everyone start to do this...
0
0
4
460
If you try to compile a list of things that cannot currently be done on Solana, you will be surprised to find nothing but a blank page.
Tech is one thing and must be good enough to unblock everything else. Solana is basically here.
GTM is the secondary component and where most products fail. It's incredibly hard and it doesn't matter where you're deployed, bad GTM can fail your product.
6
3
11
1,405
Umberto | Mostly Data 🧙♂️ retweeted
Introducing the Solana Asset Registry
Fact checking a Solana token? This is the tool for you.
A canonical search engine for every Solana asset, classified, with data from @AlliumLabs, @coingecko and @RWA_xyz behind every entry.
tokens.xyz/registry
Fact checking a Solana token? This is the tool for you.
A canonical search engine for every Solana asset, classified, with data from @AlliumLabs, @coingecko and @RWA_xyz behind every entry.
tokens.xyz/registry
54
26
237
14,839
Umberto | Mostly Data 🧙♂️ retweeted
Consolidation is happening in blockchain.
Apps are moving to infrastructure that has:
→ Handled the most traffic
→ Consistently increased capacity
→ Proven institutional-grade security
Only one network clears that bar for critical financial infrastructure.
Solana.
Apps are moving to infrastructure that has:
→ Handled the most traffic
→ Consistently increased capacity
→ Proven institutional-grade security
Only one network clears that bar for critical financial infrastructure.
Solana.
34
36
253
21,409
Slot time has been decreased to 200ms.
Solana is running at ~200ms since a while.
It was widely thought this was going to cause troubles, but that's not what happened.
We are few hours away from Alpenswitch, so it's useful to recap what happened 👇
Solana is running at ~200ms since a while.
It was widely thought this was going to cause troubles, but that's not what happened.
We are few hours away from Alpenswitch, so it's useful to recap what happened 👇
4
7
37
2,659
If we look at handoffs when moving leaders around the globe, we see that hot paths are rare and always involving South America and Asia.
Geo aware leader will help with this.
Geo aware leader will help with this.
1
0
1
143
However, vote latency for some validators went up. Still less than 2 vote latency on avg.
Not really an issue, but Alpenglow will fix this.
Not really an issue, but Alpenglow will fix this.
0
0
1
136
This is how we are going to win.
yes, but more importantly, need to reduce perf jitter to enforce tight market tick.
basically all the cracked work Alessandro and team have been doing the last few years.
must be able to differentiate between egregious stretching vs. P99 perf outlier of honest guys
0
2
12
619
There is one thing a lot of people in crypto don’t know about me: before joining the space I was obsessed with Bayesian analysis.
One concept in particular stuck with me: dependency on the prior.
This trained me to always ask “what happens if”.
Since then I’ve tried to measure how much my conclusions depend on prior assumptions, and to update those priors when due.
Prior updating is a very underrated concept.
I find it the only way to actually understand a topic instead of just confirming what I already believed.
Anyone should practice prior updating.
That’s what I did after reading this. I updated my prior and checked for what I wasn’t looking at (because I'm dumb and I wasn't thinking hard).
Yes: ~14% of the transactions from this informed searcher (the one using knowledge of future orderflow) buy after the trigger and close without any user transactions in between.
14%.
One concept in particular stuck with me: dependency on the prior.
This trained me to always ask “what happens if”.
Since then I’ve tried to measure how much my conclusions depend on prior assumptions, and to update those priors when due.
Prior updating is a very underrated concept.
I find it the only way to actually understand a topic instead of just confirming what I already believed.
Anyone should practice prior updating.
That’s what I did after reading this. I updated my prior and checked for what I wasn’t looking at (because I'm dumb and I wasn't thinking hard).
Yes: ~14% of the transactions from this informed searcher (the one using knowledge of future orderflow) buy after the trigger and close without any user transactions in between.
14%.
2/ in 86% of rounds, real users buy the same token between the bot’s buy and its sell. with no users in the window it loses ~35 mSOL. the more SOL users put in, the more it makes, up to +0.43 SOL a round. the losses don’t make it harmless, the users are the profit
3
1
25
1,554
Just to make it clear: this points again in the direction of “not trading on the knowledge of future orderflow”.
The updating the prior thing was a joke…🫠
The updating the prior thing was a joke…🫠
1
0
1
156
Now, how can a "trader with inside knowledge of future order flows" place orders, knowing that no one will buy subsequently?
2
0
3
290
1/8 There are rumors of a new kind of sandwich on Solana.
However, a sandwich needs one thing: knowing your trade before it lands.
We looked at the on-chain data and found no sign of that. What we see is a momentum bet.
👇🧵
However, a sandwich needs one thing: knowing your trade before it lands.
We looked at the on-chain data and found no sign of that. What we see is a momentum bet.
👇🧵
27
30
151
27,542
7/8 The cherry on top: we watched the same tokens before the big buy (we don't have a predicting model, but if the searcher has, the token is the one targeted...).
Every buy of similar SOL size in the previous ~2 minutes is followed by just as many buyers within 3 slots as the buy → more buys” pattern is already there. It’s public. Anyone could bet on it.
Every buy of similar SOL size in the previous ~2 minutes is followed by just as many buyers within 3 slots as the buy → more buys” pattern is already there. It’s public. Anyone could bet on it.
1
0
17
1,512
8/8 To sum up:
• It loses ~45% of the time
• It reacts to a big buy that is already public
• It sells on a timer, while buyers keep coming
• The same pattern shows up on the token without it
The only advantage here seems the knowledge of the trigger, which the bot never frontrunned.
No sign it knows the trade flow before it lands. That’s pointing to a momentum bet, not a sandwich.
Calling something a sandwich needs evidence the bot knew the flow. Here, the data says it likely guessed.
• It loses ~45% of the time
• It reacts to a big buy that is already public
• It sells on a timer, while buyers keep coming
• The same pattern shows up on the token without it
The only advantage here seems the knowledge of the trigger, which the bot never frontrunned.
No sign it knows the trade flow before it lands. That’s pointing to a momentum bet, not a sandwich.
Calling something a sandwich needs evidence the bot knew the flow. Here, the data says it likely guessed.
2
1
27
2,229