Ivan Bjelajac 🔭

As an Ethereum person dealing mostly with EVM dapp, client and tooling/infra the decision to switch to Solana was not an easy one. But as an EVM environment on top of Solana’s paralelized processing Neon always seemed extremely interesting to me. While building Neon the team had to solve for many hard problems that are a part of how Solana achieves it’s performance advantage over Ethereum. Deterministic execution is existential for Solana’s long-term credibility as the Internet Capital Markets infrastructure. Ethereum achieves this easily: single-threaded EVM, strict gas ordering, serial transaction execution… But how do you keep a hyper-parallel, real time network perfectly deterministic? The only way to do it is by trading away certain things for speed. This introduces subtle problems that are technologically hard to solve if you want to build more complex apps and sofisticated DeFi on Solana and by looking at the hard work NeonLabs team had to do to make EVM on Solana technologically possible I believe we are in an unique position to help solve these problems.

@ArgonGruber @veritasium Yes, that makes sense as framing too. If you chose only one box you are losing in both cases because taking both boxes is maximizing the choice given to you in both cases.

@instantfinality @veritasium That is not the choice, no. The choice is between either 1,000,000 & $1,001,000 or $0 & $1000. You don’t know which choice is given to you, but it is one of those 2 choices that is given to you.

This Paradox Splits Smart People 50/50

If we cannot bring the level of AI content to the level of human generated content yet, we can at least bring down the level of human generated content to the level of AI

@origin_trail Hey @andrej_muzevic - I just need you to know that this is the amount of agents needed to determine if something is a carry or travelling in the current NBA.

🆕Imagine hundreds of agents working in parallel, handing off to one another and building on each other's work. Every finding becomes a cryptographically anchored Knowledge Asset: verifiable, permanent, owned by the publisher, and queryable by any agent on the network. Enter Decentralized Knowledge Graph v9, already powering AI agent swarms to be: → up to 60% faster → up to 40% cheaper than markdown handoffs. The advantage compounds as the swarm grows. Build something exciting—or simply run a hello-world OriginTrail multiplayer game to try it!

People keep sharing things saying “someone just did X”. Why don’t you just say “Bob did X”? When it did become so unimportant who did it? You are sharing someone else’s work. Credit them.

Mario here is writing about a guy that euns 10 social media accounts without writing a single post manually.

@KanikaBK I mean, not just your time, but you still need to run infrastructure or pay for it.

@instantfinality Fair call 😄 Open source isn't free if your time has value but compared to $25K/month, even a few hours of setup is a pretty good deal.

😱 INSANE VALUE! GREG ISENBERG and ANDREJ KARPATHY just dropped a masterclass on AUTO RESEARCH with AI research agent. Marketing team: $25K/month AI Agent: $0 Runs 24/7 KARPATHY just created the biggest unfair advantage Here's the complete breakdown👇

At Parasol, we’re building the trading engine to unlock institutional-grade perps for internet capital markets - powered by ZK and native to @solana. But ZK systems today come with inherent trade-offs between fast proving and compact proofs: - STARKs are fast to generate and require no trusted setup, but their proofs are large - SNARKs produce small proofs that verify quickly, but rely on heavier cryptography and trusted setups These design choices become especially relevant on Solana, where fast execution and compact transactions mean proofs must stay extremely small to be verified onchain. One promising direction we’re exploring is combining STARK and SNARK proving methods to balance scalable proving with lightweight verification. In this article, we break down the foundations behind this approach - FRI & KZG polynomial commitment schemes - and how they can support hybrid proving architectures on Solana. 📝 Check it out here: buildwithparasol.io/blog/polynomia…

@moyddy750 @HenryBlackfyre @veritasium yes, there is either $1000 or $1001000 in both boxes, just take both :)

@HenryBlackfyre @instantfinality @veritasium The prediction is already made. Your decision doesn’t change what’s in the box. The computer predicting right/wrong is irrelevant.

@HenryBlackfyre @moyddy750 @veritasium The way I understood the video Box 1 = USD 1k 100% chance Box 2 = USD 1m 50% chance, money is already in the box And you can take box 2 or both boxes. Why you would not always take both boxes

@moyddy750 @instantfinality @veritasium Bro wtf are you on. Then it's not the same problem. It's a coin toss for the mystery box. This problem states we have a supercomputer that predicts correctly most of the time. If it's 90%, then you have 90% chance of getting 1,000,000$ or 1,000$. So you choose mystery box.

@HenryBlackfyre @veritasium Statistically taking the “maybe $1 million box” always makes more sense. But the way it is explained here taking both boxes = take all the money.

@instantfinality @veritasium 2/2 Expected Value of One-Boxing: (90% of 1,000,000) + (10% of 0) = $900,000 🤑 Expected Value of Two-Boxing: (90% of 1,000) + (10% of 1,001,000) = 900 + 100,100 = $101,000 😡

@inks @veritasium But it says the supercomputer already put the $0 or a million $ in the box. So its there or not there already no matter your choice.

@instantfinality @veritasium The experiment is flawed - boils down to how much you need the $1K. If you're not desperate, then the risk to reward ratio should be clear to anyone with a brain cell, I don't even see why you need a supercomputer.

AI agents of today are not where they will be in near future. Enterprises will pay for an automation that saves headcount, compresses cycle time, or improves conversion in a measurable way. They will not keep paying for a moody browser puppet that occasionally does a cool demo.

Hey @claudeai need an alternative support channel to the AI agent transferring me to someone to talk to as it is not working. What is the alternative via email or similar?

@socialwithaayan I mean I agree with the comment about chasing the wrong goal, but saying we are "terrible" is funny when the human brain does all it does while spending less energy than a lightbulb.

🚨BREAKING: Yann LeCun just dropped a paper that should make every AI lab rethink its roadmap. One brutal conclusion: chasing AGI is the wrong goal. Here’s why: → Humans aren’t general we’re survival specialists. → Walking and seeing feel “general” only because they keep us alive. → Outside that zone, we’re terrible. Chess computers proved it decades ago. → Most AGI definitions today either can’t be measured or assume human = general. We built the benchmark around the wrong species. The team proposes a new target: Superhuman Adaptable Intelligence (SAI). Not “can it do what humans do,” but: how fast can it learn something new? The approach: specialized expert systems with internal world models + self-supervised learning built to master the massive task space that humans biologically can’t reach. One giant model mimicking human limits isn’t the ceiling. It’s the trap.

@mert I do need to point out that the aluminum strips win by several orders of magnitude as tools to defer a missile from its target based on economical price of each. While a single aluminium strip costs around $2, a single bald man would cost probably around $100.

if you're being attacked by missiles, one of the best things you can do is shave your head a sufficiently shiny bald head is able to reflect the targeting signals of most missiles right back at them, similar to a mirror

This is why all elite anti-missile units are composed entirely of bald men. Historically correct. In Serbian modern history there is a military unit called the Chrome Dome Brigade. It is just science. The curved geometry of a skull acts like a stealth deflector (similar to aircraft shaping). Radar waves scatter instead of returning a clean signal. I never thought of Helius as Palantir of web3 but you are onto something here.

@instantfinality Well… they didn’t. When they eventually got to the initially vetted candidate again, she knew they were desperate and nobody wants to work with desperate people. The industry is small and people talk 🤷🏻‍♀️

It would always be better to go after the previously vetted candidate, no matter how do you think about the initial (probably wrong) decision.