Research @SolanaFndn PhD in Physics - Cosmology - Solana Views are my own.
- Tweets 2,314
- Following 1,578
- Followers 1,733
- Likes 3,079
Is @RobinhoodApp still a thing?
🚨BREAKING: @Solana has flipped @RobinhoodApp to become the No. 1 chain for tokenized commodities spot DEX volume, with a 45% market share.
Just three weeks ago, Solana held only 1% of the market.
6
276
I’m shocked. I’m so sorry to read all of that…
I’m sure @vibhu is going to be ok soon, and he will be happy to know you bought a ticket with a 25% discount using BP26-UMBERTO
I was recently diagnosed with an incurable and rare disease.
I've sat with that sentence for weeks, trying to find a way to say it out loud.
I was having dinner in London with friends and somewhere in the middle of it I felt myself leave the room and it just never went away.
I ended up seeing a specialist. When she gave me the diagnosis, she said it plainly and let it sit there.
She said there's no cure. She said the only thing that helps is being around others who have it. That the isolation is worse than the condition.
I've thought about that a lot. About how long I've spent alone with something I assumed no one else understood.
But I'm not going to do that anymore.
The disease is called Breakpoint.
Use code BP26-VIBHU to get 25% off, and if you DM me today I'll also send you $100 USDC.
1
9
1,775
I’m sure that no one really reviewed these plots (produced by their best data scientist…).
At least they removed the scientific research performance.
AGI is here, our job is at risk.
At its default effort setting, Opus 5.5 delivers frontier results for a fraction of the cost per task, often beating other models running at their highest settings.
It also generates output more than 30% faster than Opus 5.
1
205
I think no-panic needed.
"I think shorter slots make forking more chaotic" here just means that slot time reduction makes it more visible, involving more subsequent slots.
But skip rate is still incredibly low, with no change post slot time reduction
the short answer is forking. This used to happen all the time now it's so rare it's kinda scary when it happens lol
34, 35, 36, 37 were skipped (whole round)
GREEDkgav1ox1jYyd9Anv6exLqKV2vYnxMw5prGwmNKc built 38, 39, 40, 41 on top of 33. It was slow to build them, likely destabilized from the previous leader skip (this is a known issue), so 40 and 41 were actually skipped
Fc6NNdS2j3EmrWbU6Uqt6wsKB5ef72NjaWfNxKYbULGD built 40, 41, 42 and 43 on top of 39. It was also kinda slow to produce.
BSVckjdW2f8kcXPGcrPPtV9kUDBZ8w8PjrrGVnxgEdwq built 44 on top of 43. From here things recovered, 44, 45, 46 and 47 were fast BUT for some reason - maybe a network issue, maybe a turbine issue (investigating) - 44 took 1.5s to propagate over turbine
after that things stabilized
I think shorter slots make forking more chaotic
1
12
953
I think this is more what @cavemanloverboy would say, but fine
Welcome to Solana High.
Class starts Thursday at 2PM UTC sharp.
Sign up to the @ecosystemcall below.👇🏼
1
1
7
590
This, my friends, this is complete retardness.
AI can help you, but it’s just another tool. Who believes we can already replace anything is lying to you.
The truth is that the quality of the accepted outcome is declining.
This is why “it seems” AI is threatening something; we are basically not willing to think anymore, so in this framework AI can replace whatever.
2
1
8
574
You are not ready for what v1 tx format unlocks.
You are not bullish Solana enough!
This is the same type of tx, same signer.
First (on the left) is v1 format.
Second (on the right) is Legacy format.
The tx is now ~50% more efficient in terms of compute and ~17% more efficient in terms of overall resources.
Do you know what are the missing CU?
4
9
58
6,840
Boys, let's upgrade to be men.
Be prepared for SBPFv3.
SBPFv3 is the new bytecode format, ready to integrate into your Solana programs now.
SBPFv3 resolves syscalls at compile time, enforces a strict ELF layout, and aligns Solana with upstream eBPF.
Solana no longer maintains a bespoke bytecode dialect.
Down the road, SIMD-0500 makes SBPFv3 the only deployable format. Its feature gate ships in the next Agave v4.4 release, so the best time to migrate is now.
The toolchain is ready and rebuilding is a few commands for most Rust programs.
Full migration guide: anza.xyz/blog/migrating-sola…
1
2
21
1,405
six years of workarounds, will soon be deleted
Btw, soon means “be ready”
fun fact: solana's transaction size limit was never a design decision. 1,232 bytes is just what fits in one IPv6 packet.
transaction v1 will be live soon. 4,096 bytes. ZK proofs and large multisigs in a single atomic transaction.
six years of workarounds, will soon be deleted.
1
3
18
2,794
We can do the following experiment.
We can take 2 validators, Validator A and Validator B, which seems to adopt 2 different block building pipelines but same client version (Firedancer).
We can show how Cost Unit affects execution, and we discover the 2 correlates: more Cost Units implies higher execution time.
But now you start to look deeper, and you notice few differences. The first is the variance. Despite correlation is pretty much the same, you see that one validator has slightly higher variance for execution time, on par of costs.
So, you start to feel that there are implementation details affecting execution time.
To check this, you fix the sample to be vote txs at 3,465 cost units and compare the two validators. You see that execution time (for exactly the same type of txs) is different.
We finally compute the wall-clock execute time over slot duration. We discover that again we see a difference, a non-negligible one.
Validator A averages at ~18%
Validator B averages at ~31%
Thus you can't conclude that Cost Units is a good meter for "consumed time".
Validator A burns 18% of the slot executing. Validator B burns 31%. A fee that prices "time" via CU is pricing a unit that changes with the implementation.
nitter.net/MostlyData_/status/209…
> You can clearly see the auctions and the white space is extra capacity.
At this point, wondering why someone should calibrate base fee on something that “consumes time” if there is a fair portion of slot time where the validators just wait.
More I work on a fee model, more I’m convinced that this doesn’t make sense.
And to convince you, I have a 50 pages analysis…
7
3
22
14,803
> You can clearly see the auctions and the white space is extra capacity.
At this point, wondering why someone should calibrate base fee on something that “consumes time” if there is a fair portion of slot time where the validators just wait.
More I work on a fee model, more I’m convinced that this doesn’t make sense.
And to convince you, I have a 50 pages analysis…
Please enjoy agave with rpc turned on and replaying mainnet on a single replay thread at 35% use.
You can clearly see the auctions and the white space is extra capacity.
Since we have the base for a config file now, someone should send a PR to extend the config/tune it for non voting validators.
Full agave on 8 cores, would be a very cool demo!
2
12
15,930
STOCKLANA
The stock market is open for building.
One week to build something innovative with stocks on Solana. $100K in prizes.
Sept 11 - 18, opening bell to closing bell:
hackathons.solana.com/hackat…
486
343
2,323
616,631