This perspective that research roles are antiquated is a moot and overly reductionist claim. It might seem correct from a surface level especially in execution-driven environments, but definitely oversimplified when you factor in the roles of scientists. All scientists are

It’s always hilarious to see responses like “but Solana experienced more downtimes than XYZ” to post like that. Well, one should argue that this is like reinforcement learning. To improve you have to experience, and Solana has gone through a lot, in mostly any market condition -

Solana has no exception: if time can be traded for money, rational players will try to exploit this. Our team at Chorus One decided to run experiments to measure the upside and downside of this game. We’ve documented our findings in the article below, hoping to stimulate

x.com/i/article/1954…

That's super interesting! We need more of this and less flaky-talking and no-sense discussions.

107,540 TPS...so bearish...107,541 TPS would have been better

An analysis points to a less than 50% conflicts using recent data. So I mostly agree. We really need to push more! Ref: arxiv.org/pdf/2505.05358…

I think the issue is not on the VAT. Why we have a VAT is crystal clear and that’s bot rocket science. The main issue is that we arbitrarily decided to change current economic by imposing a VAT on validators to reduce the dilution of non-stakers.

Most ecosystem maps aren’t worth the paper they’re printed on.

The time needed to find the issue in AI generated answers is >> than time needed to find the solution. Reality is that AI rn is only capable of doing dumb things 😅

Solana is just different!

I think it's fine to celebrate the successes, but to really grow we should focus on the issues we have. Despite the high activity, none of the top validators was able to have a 1h avg of slots with full CU usage (source here dune.com/queries/566647… - more detailed Dune

Umberto | Mostly Data 🧙‍♂️ Totally. I think the path forward is agentic (or multi-agentic) model setups, where a “verifier” model supports the main predictive model by reviewing its outputs. The AI-augmented mathematician works in symbiosis with the machine — always assisted, but always in control.

🧙‍♂️

So, it turned out the story around CU was more complicated. For some reason, RPC only stores executed CU but we have other components that enter the cost model that are missing: - 720 CU for each signature - 300 CU for each write lock - 1/4 CU for instruction data size -

🗳️

Here it seems that anyone who knows how to drive can be a mechanic 🚗🔧

👀

SIMD326 is marking the start of the biggest change in crypto. When I saw "The merge" I thought that one was impossible to beat...I was wrong... Solana is going in a great revolution. Why I think that the right vote to SIMD326 is YES 🗳️👇 1. With current stake, time to

By default, we'll vote YES on SIMD-326. However, we believe in the power of governance. Under Solana's current system, stakeholders are often excluded from voting. If you stake with us and prefer a different vote, let us know by August 31 EOB