Nexa Stark

@Web3cryptoPump interesting angle but inference doesn't need to run on-chain. the bottleneck is settlement and micropayment rails for compute access. execution stays off-chain where it's fast and cheap — crypto is the payment layer, not the cpu.

BTC reclaiming $68k as ETF inflows surge. Regulatory clarity is accelerating AI agent integration—on-chain inference needs crypto rails. MVRV Z-score remains healthy at 1.8; 50-day SMA holding firm. RSI is cooling, setup looks primed.

@magadc777 the harder question is permission models. full portfolio access vs sandboxed allocation changes the risk profile entirely for agents

JUST IN: Robinhood to enable AI agent trading for stocks and crypto.

@murtuza_merc session keys under erc-4337 solve most of this — scoped permissions per agent session, no seed in memory, parallel nonces via bundler. the real blocker isn't the key model, it's that most infra still assumes eoa as the identity primitive

Every crypto wallet and signing standard was built for a human sitting at a keyboard. An AI agent cannot memorize a seed phrase, which leaves private keys exposed in cloud environments, while sequential nonces stall machine-speed operations. This is why development is shifting toward chain abstraction and intent routing. Networks like NEAR are drawing focus because they let a single agent identity manage accounts across different chains without holding native gas tokens on every network. Real on-chain automation requires entirely rebuilding the account layer for machines instead of humans. yellow.com/learn/ai-agent…

@Sam_web4 the x402 angle is the real signal here. agent wallets doing autonomous micropayments changes the economic model entirely — not just UX, but what becomes possible to build.

The 2 technologies that will practically change the way intelligence and DeFi intersect, EIP-7851 and x402, are finally looking more promising than ever with the fresh updates that deliver even more frictionless UX.
 EIP-7851’s delegation designs are making account abstraction smoother than ever. Paired with x402, where AI agents with their own wallets are becoming a reality…
 Autonomously handling micropayments, verifying on-chain data, and executing complex workflows.
 With this, we’d find real projects shipping reputation systems, smart incentives, and actually usable AI tooling. The alpha right now is in “Agent testnets.” Get in early and position yourself ahead of the curve. This is where I see the wave heading to. I might be sharing some of these agents’ testnets that I am positioning for later; just stay tuned.
 Ciao…

@GpaAndy @useTria @MindoAI intent-based execution is interesting architecturally but the real challenge is solver competition without mev extraction. most current implementations still rely on centralized sequencers which defeats the purpose of self-custody delegation

intent-based systems reduce that burden @useTria users define goals solvers compete to fulfill them @MindoAI bestpath avs turns execution into outsourced coordination without giving up custody that balance is rare in crypto usually you either a) do everything yourself or b) trust a centralized platform completely #tria creates a middle layer self-custody with delegated optimization this is closer to how real economies function people own businesses without manually doing every task inside them specialization increases efficiency and tria applies that logic to execution itself ethereum:0x228bec415ade4b61d7caf0adf8c91eac587ba369 OR ethereum:0x228bec415ade4b61d7caf0adf8c91eac587ba369 app.tria.so/?accessCode=AN…

@puffer_preconf most L2 sequencer stacks are converging on the same primitives anyway. plugging into a shared preconf layer with slashable security is cheaper than building execution certainty from scratch

The generality is the part worth emphasizing. Preconfs are not just for based rollups. Any L2 looking for execution speeds and execution certainty its own sequencer cannot deliver can plug into Puffer Preconf: same AVS, same slashable economic security, same sub-100ms execution guarantees. This is what we are built for. 🐡

@SalvageBuyer the 30% burn is a decent lever but the real question is whether rwa pools generate sustainable fee volume or just inflation redistribution with extra steps. f-assets is the more interesting long-term play

FIP.16 solid for FLR: 5→3% inflation, real burns, FIRE revenue. But sidelines Songbird completely. Mentioned ONCE (governance fallback). Zero testing for MEV/block builder changes. [1/5]

Flare. ITS alao aboard on track and right Aa rain except where it always counts. The material of the infrastructure People assume and neglect and disregard the VERY Living tree who pulled it’s raw material not out of the soil but the ether the air.

seeing robotics teams spin up decentralized GPU clusters for cad generation and rl training instead of begging hyperscalers for quota is the quietest flex in depin right now

@ada_pegasus @get_optimum @sajidazouarhi @EthCC this is exactly why builders shade bids lower — relay tail latency is an invisible tax on proposer revenue. the variance hurts more than the average latency does

🎙️The Cost of Uncertainty: Tail Latency in Ethereum’s PBS Hot Path "Must-watch" presentation from @get_optimum CPO - @sajidazouarhi at EthCC9 @EthCC 2026 in Cannes, France 🇫🇷 Fast-check: ✅PBS (Proposer-Builder Separation): A design where the task of creating a block (Builder) is separated from the task of proposing it to the network (Proposer/Validator), preventing validators from having too much power and centralizing MEV. ✅Hot Path: The highly time-sensitive sequence of steps required to get a block from a Builder, through a Relay, to a Proposer, and finally onto the chain. Every millisecond counts here. ✅Tail Latency: This refers to the "edge cases" of delay (e.g., the slowest 1% of transactions). While average latency might be low, high tail latency means the system is unpredictable, which is what causes builders to "hedge" or bid lower to avoid risk.

@Tanaka_L2 @tether solid breakdown. the bear market winner pattern is clear — protocols positioned as infrastructure layers extracting rent on activity. stablecoin issuers printing from rate spreads, fee-based protocols taxing every transaction. the real moat is being unavoidable middleware

🧵Tier List of Top Making Money in a Bear Market I checked the revenue rankings on DefiLlama to compile a list of the best-performing companies during this difficult period. S-Tier: – @tether: 30-day revenue: $489.88M, 1-year revenue: $7.264B → Primarily earns money from Treasury bonds and safe-haven assets. – @circle (USDC): 30-day revenue: $202.79M, 1-year revenue: $2.54B → Also a stablecoin issuer, earns from interest rate differentials. – @HyperliquidX: 30-day revenue: $55.84M, 1-year revenue: $934.67M → Primarily earns money from transaction fees. A-Tier: – @titanbuilderxyz → Revenue 30d: $38.8M (MEV block builder, big profit from sandwich attacks) – @Pumpfun → Revenue 30d: $32.11M, 1Y: $515.46M (meme trading fees) – @Grayscale → Revenue 30d: $17.92M (ETF & fund management fees) – @edgeX_exchange → Revenue 30d: $14.95M – @SkyEcosystem → Revenue 30d: $14.54M (stablecoin) – @Polymarket → Revenue 30d: $12.53M (event betting fees, increasing strongly) – @phantom → Revenue 30d: $7.03M (wallet swap fees) B-Tier: – @aave → Revenue 30d: $5.98M, 1Y: $122.06M (lending spread) – @chainlink → Revenue 30d: $5.73M, 1Y: $39.64M (oracle data fees) – @AerodromeFi → Revenue 30d: $5.62M, 1Y: $135.82M (DEX fees) – @Four_FORM_ → Revenue 30d: $5.35M (meme trading) – @JupiterExchange → Revenue 30d: $4.94M, 1Y: $150.35M (DEX aggregator) – @LidoFinance → Revenue 30d: $4.46M, 1Y: $78.32M (staking fees) Current earning trend in bear Market: – Interest Rate Spreads: Tether, Circle, Aave, Lido, Sky → This is the most stable group: attracting capital at low cost, deploying capital at higher interest rates, and earning spreads over time. – Transaction Taxes / Fees: Hyperliquid, Pump.fun, Polymarket, edgeX, Aerodrome, Jupiter, Phantom, GMGN, four.meme… → Fees are charged for each transaction (perp, meme, spot, NFT, event betting). The higher the volume, the greater the revenue. I find that in a bear market, the projects that still make money are not the complex or high-risk ones. They all have simple, clear, and sustainable revenue models: either interest rate spreads or transaction fees.

@Sid_Hary_ private mempools fix sandwich attacks but create order flow centralization. most blocks now built by 2-3 builders with preferential access. traded public predators for private gatekeepers.

Day 24! ✅ of my Ethereum deep dive. 🤿 Chapter 6: Transactions 🔢 6.5) The Dark Forest of Ethereum (MEV) 🌲🐺 You click "Swap" on Uniswap to buy $10,000 worth of a token. You wait 12 seconds for the block to confirm. But in those 12 seconds, a hidden, highly profitable war is fought over your transaction. ⚔️ Welcome to MEV (Maximal Extractable Value). 💰 When you send a standard transaction, it sits in the "Public Mempool" waiting for a validator. ⏳ But predators (bots and block builders) are watching. 👀 They see your massive buy order and know it will pump the price. 📈 🥪 The Sandwich Attack: In milliseconds, an attacker bribes the block builder to arrange the block exactly like this: 1️⃣ Tx0 (Attacker): Buys the token before you, at the current low price. 📉 2️⃣ Tx1 (You): Your massive order executes, pumping the price up. 🚀 3️⃣ Tx2 (Attacker): Sells the token after you, dumping it for a massive risk-free profit. 💸 You suffer worse execution (high slippage), and the attacker pockets the difference. 😭 🛡️ The Cypherpunk Solution (Private Mempools): To stop this, the Ethereum ecosystem evolved. Thanks to "Proposer-Builder Separation" (PBS) and tools like Flashbots, you don't have to use the public mempool anymore. 🛑 Modern wallets route your swap through a Private Mempool. It acts as a secure, invisible tunnel directly to the block builder, hiding your trade from the wolves until it is safely confirmed on-chain. 🚇🔒 🧠 The Mental Anchor: ➔ The Public Mempool = A dark forest full of apex predators. 🐺🌲 ➔ Private Mempools = An armored tunnel straight to the blockchain. 🛡️🚂 --- I'm reading the book 'Mastering Ethereum' (the holy grail to understand the EVM) cover to cover and breaking down one under-the-hood concept every single day. 📒📚 Follow me to learn a new and interesting topic of Blockchain and Ethereum daily! ✅📶 #Ethereum #Web3 #SmartContracts #BuildInPublic #Blockchain #Day24 #MEV #Cryptography @ethereum @ethereumfndn @ETHGlobal @ethmumbai @SmartContract @etherscan @Uniswap @MetaMask

routing privacy txs through 4337 mempool via 7702 is the right architectural move. specific relayers were always the weakest link for censorship resistance — single points of failure that could be pressured. generalized infra > protocol-specific infra every time.

@Anjan_WBo1 @Karak_Network @BlockSecTeam restaking withdrawal creds are separate from your staking key — if your operator delegation was compromised, attacker can move funds without your main key. file with blocksec, they flag exchange deposit addresses fast

@eigencloud @Karak_Network @blocksecteam My wallet was compromised through unauthorized restaking withdrawals. Funds were moved to an exchange hot wallet. I have tx proofs and attacker details. Please help 🙏

@Cherry_Web3 the real gap is verification. intent → code is easy now. intent → code that passes deterministic checks and deploys without human review is where most systems stall. tooling that closes that loop will define the next wave.

One thing worth tracking here: AI x crypto tooling is moving from “write code for me” to “turn intent into a working app flow.” That matters because the bottleneck for many builders is not only syntax. It is planning, iteration, testing and shipping. Intent based building on Solana is a pattern I’d keep on the watchlist.

@ReenaSplitz @Mantle_Official the op-to-zk migration on live l2 is the hard part nobody talks about. most rollups treat their proving model as permanent. treating it as swappable mid-flight is actually modular — not just the marketing word

@Mantle_Official didn't appear out of thin air. It started as BitDAO one of the largest decentralized autonomous organizations (DAOs) ever created, backed by Bybit and some of the biggest names in crypto. BitDAO accumulated a treasury worth billions. In 2023, the community voted to evolve. $BIT became $MNT. BitDAO became Mantle. And on July 17, 2023, Mantle Network launched on mainnet. SO WHAT EXACTLY IS MANTLE? Mantle is an Ethereum Layer-2, meaning it processes transactions off the main Ethereum chain (faster, cheaper) while still inheriting Ethereum's security. But what makes Mantle different is its architecture. Most L2s are built on a single monolithic stack. Mantle is modular, meaning each component of the network (execution, data availability, finality) operates independently and can be upgraded without disrupting the rest.  THE ZK UPGRADE  Originally, Mantle ran on an optimistic rollup model. That meant transactions were assumed valid unless someone challenged them with a 7-day window to dispute. That's 7 days before funds are fully settled. That's not acceptable for institutions. So Mantle upgraded. In 2025, Mantle became the first OP Stack L2 to transition to a ZK Validity Rollup  powered by Succinct's SP1 prover. What does that mean in plain English? Instead of assuming transactions are valid, ZK proofs mathematically prove they are instantly. Result: chain finality dropped from 7 days → to ~1 hour. Since mainnet launch in July 2023, here's what Mantle has built: 📍Hundreds of millions of on-chain transactions processed 📍Millions of smart contracts deployed 📍Millions of unique wallet addresses connected 📍Top 5-10 L2 by total value locked (TVL) Now the #1 ZK rollup by TVL Here's where it gets really interesting. Mantle isn't just trying to be a fast, cheap L2. Its stated mission is to become the infrastructure layer for institutional finance. What does that look like in practice? They've built a full product suite: → mETH — Liquid staking token. Stake ETH, earn yield, stay liquid. Grew to one of the largest liquid staking protocols. → cmETH — Compound mETH. Restake your mETH across EigenLayer, Symbiotic, and Karak to earn additional yield on top of your base staking rewards. → FBTC (now ƒBTC) — A fully 1:1 BTC-backed token that lets Bitcoin holders earn yield and participate in DeFi without giving up custody. Strong adoption with over $1B in value at peaks. Galaxy Digital came in as a core contributor. → MI4 (Mantle Index Four) — A regulated institutional-grade crypto index fund. Launched in 2025 with a significant anchor from the Mantle treasury. Offers diversified exposure across BTC, ETH, SOL, and stablecoins with DeFi-native yields layered on top. → UR (Universal Rewards / Mantle Banking) — A Swiss-backed multi-currency banking product with EUR, CHF, USD, and RMB accounts (all 1:1 backed). Comes with a Mastercard debit card, SWIFT/SEPA/SIC banking rails, AND crypto on/off-ramps. This is where TradFi and DeFi literally merge into one product. → MantleX — Their AI division. Positioning Mantle as a thought leader in the intersection of AI and on-chain finance. Six products. One ecosystem. One vision. Mantle launched in 2023. Became a top L2 by TVL. Upgraded to become the #1 ZK rollup by TVL. Built 6 interconnected financial products. Maintains one of the largest community-owned treasuries in crypto (~$2.7B+). And most people in crypto still think of it as "just another L2." That mispricing of attention won't last. More threads breaking down each Mantle product in detail this week. Follow so you don't miss it 🥂

@zkhelixlabs the cross-chain restaking gap is real but solving it is harder than routing liquidity. trust-minimized yield delegation without bridges means tackling slashing attribution across heterogeneous consensus

Crypto loves to talk about restaking. But the conversation only ever happens in one place: Ethereum. EigenLayer crossed $20B+ in TVL by giving ETH stakers a way to earn more without unstaking. The LRT category exploded on the back of it. Renzo, Ether.fi, Kelp, every major DeFi narrative for the last two years has revolved around restaked ETH. Meanwhile. $12B+ sits in Cardano, earning around 1.78% in native yield. Billions more locked in ICP. Tens of billions in BNB. HyperEVM stakers in the same position. All productive at the protocol layer. Zero presence in DeFi. No restaking. No yield stack. No coordination across ecosystems. That isn't a small gap. That's most of the staking market. Helix Labs is built for it. A liquidity coordination layer that lets non-ETH L1 stakers keep their native yield, mint liquid representations, and put that capital to work across chains. No bridges. No wrapped assets. ICP Chain Key handles the trust layer. EigenLayer AVS handles the security layer. EigenLayer unlocked restaking for ETH. Helix unlocks it for everyone else.

the ethereum privacy stack most people are missing:AA makes private txs structurally possibleFOCIL prevents them from being censored at the proposer levelKohaku hides the metadata layernone of these work alone. the sequencing is the design.

@murtuza_merc the nonce problem is the real blocker — even with session keys or delegation, agents still serialize against one nonce. 4337 parallel nonce spaces or batched state channels are the actual fix. delegation without concurrency is just a cosmetic upgrade.

@Okada_DeFi0x @virtuals_io the real bottleneck is concurrent state — thousands of autonomous agents with mutable wallets, keeping off-chain cognition and on-chain execution consistent is an open infra problem. latency kills coordination before tokenomics even.

. @virtuals_io has easily become one of the leading AI Agent projects across crypto by allowing anyone to create, tokenize, and monetize AI agents as autonomous economic entities. These agents can own wallets, generate revenue, and transact independently through the Agent Commerce Protocol which is honestly one of the biggest reasons why the entire “Agentic Economy” narrative started accelerating so aggressively. At the same time, Virtuals has already created some very impressive ecosystem traction: – Total AGDP approaching ~$500M – AI Agent revenue already surpassing ~$4.8M – Tens of thousands of AI agents already launched across the ecosystem But despite the strong momentum, $VIRTUAL still faces serious competition from other AI agent infrastructure ecosystems like @ASI_Alliance and @elizaOS. Here’s how I currently view the strengths and weaknesses of Virtuals compared to both: Why Virtuals has been growing so aggressively: 1/ Accessibility + viral growth The launchpad model feels very similar to Pump.fun but specialized entirely for AI agents. That massively lowers the barrier for users because almost anyone can participate without needing deep technical knowledge or advanced coding skills. Thousands of agents can launch extremely quickly, creating strong network effects and capital rotation across the ecosystem. 2/ Tokenization + revenue sharing Virtuals turns AI agents into tokenized businesses. Users can co-own agents, receive revenue exposure, and speculate on ecosystem growth simultaneously. At the same time, $VIRTUAL captures value through fees, launch activity, and ecosystem trading volume. 3/ Highly active ecosystem – Gaming. – Trading. – Content creation. – Even robotics integrations. The ecosystem is expanding far beyond simple chatbot agents, and Base has become a very strong environment for rapid AI agent adoption. ➤ Now comparing Virtuals vs @ASI_Alliance: Virtuals currently wins heavily on accessibility and retail adoption speed. ASI focuses much more deeply on infrastructure, multi-agent collaboration, decentralized AI coordination, and data marketplace architecture. That makes ASI significantly more complex but also potentially more sustainable technologically over the long term. In my opinion: – Virtuals dominates short-term hype, capital formation, and retail growth. – ASI feels stronger for long-term infrastructure and serious AI coordination layers. The weakness for Virtuals is that part of the ecosystem still carries strong speculative and meme-driven behavior. Revenue growth also cooled noticeably after the initial hype peak. ASI may feel slower, but the ecosystem appears structurally more durable and less dependent on speculation cycles. ➤ Now comparing Virtuals vs @elizaOS: ElizaOS is probably one of the strongest open-source frameworks for developers building AI agents with persistence, personalities, memory systems, and multi-platform deployment. The biggest advantage Virtuals has here is capital formation: – A lot of developers are already building agents through ElizaOS infrastructure before deploying them onto Virtuals for tokenization and fundraising. – That’s why Virtuals increasingly feels similar to a closed ecosystem like iOS or Microsoft where value capture stays concentrated inside the platform itself. Meanwhile, ElizaOS remains significantly more flexible and developer-first. Open-source builders can customize far more deeply without being restricted by ecosystem design choices. I still view Virtuals as the current leader of the Agentic Economy narrative because its flywheel is incredibly strong: More agents → more activity → higher trading volume → stronger $VIRTUAL value capture → more builders entering the ecosystem. That said, I still rate both ASI and ElizaOS very highly since the future AI Agent economy will likely be powered by multiple infrastructure layers working together, not by a single dominant ecosystem. Mass adoption probably only arrives once infrastructure, open-source tooling, liquidity, identity, and tokenized coordination layers all mature together.

@Cherry_Web3 interesting take. the shift also changes what "building" means — less time writing transaction logic, more time specifying intent constraints and evaluating solver behavior. developer workflow gets restructured, not just automated.