CME group lists its first compute futures contract on october 5th. They’ll be tracking nvidia gpu compute prices out to 36 months. It’s a financial instrument letting anyone hedge or speculate on the price of computation itself, the same structure oil futures use for a barrel. $BTC miners sitting on ai data center capacity are the most direct beneficiaries. They can now hedge compute revenue the way an airline hedges fuel costs, locking in a price months ahead instead of carrying full exposure to gpu market swings.

Tokenizing an asset is increasingly the easy part. The harder part is what happens to it afterwards: — Where does it sit? — How does it earn? — Where does liquidity come from? — How does an institution deploy against it without throwing away the custody and operational controls it already requires? I’ve studied the network @doppler_fi has built and this is probably the clearest way I understand what they’re going after. — $XRPL gives them the XRP-native market — @base opens broader onchain liquidity — @CantonNetwork puts them closer to institutional tokenized assets Then underneath that you have custody/operations through Fireblocks and Ceffu, institutional routes through SBI/Evernorth, and distribution through Bybit + Bitget Wallet. Personally, if I wanted early exposure to what gets built around XRP rather than just XRP itself, Doppler is one of the infrastructure plays I’d be looking at. Because I think the next phase of RWAs gets decided much further downstream than issuance. Getting an asset onchain creates inventory, building somewhere for that capital to actually go creates a market.

$NEAR is up 183% in a month and $TAO 37%. Same index, same news cycle, a 5x gap between the best and worst name. The gap is way more important than the 54% headline imo. Near and vvv pull away hard after september 15th. $WLD and $TAO barely move. The total market index sits flat the whole time. that's not "ai crypto is hot," that's two names carrying the entire sector. Did near and vvv actually do something tao and wld didn't?

Staked some of my $ETH on @WireNetwork. The yield goes into $WIRE pre-tokens while my principal stays liquid. Pretty innovative way to participate. Setup was simple too. Ref link > connect wallet > stake > done. ICO season might be back. ➡️ hub.wire.network?grsf=sherif…

Disappear from Lazarus. Enter the Alcove.

USDe is starting to have a life outside @ethena. Meridian putting perps + prediction markets behind the same USDe collateral on Robinhood Chain is another small example of that. Ethena doesn’t need to predict whether people want to trade SOL, SPY, gold or the outcome of an event. Someone else builds the market, USDe gets pulled into the activity. Honestly, this is a kind of distribution model I’d keep watching. The strongest monetary assets tend to become useful in places their issuer never designed for.

Agents can now open Aqua positions for you on @1inch through MCP. You give it a pair, it comes back with a range and an opening price, you sign. Everyone's focused on the keys staying in your wallet. fine. but your tokens only move when a resolver fills, which means you've approved a contract to pull them. So i'd look at that approval before anything the agent says. spender, amount, whether it's unlimited, and if closing the position actually clears it. And check the range yourself: a sloppy one just turns you into exit liquidity at a worse price. How are you checking what an agent hands you to sign?

Uptober is officially here! My conviction bags are 👇 - $ENA - $HYPE - $INJ - $ONDO - $PENDLE Shill me some mid/low caps to diversify 👀

Standard Chartered putting a $2 target on $ENA will get the headlines. But the assumptions behind it are much more useful imo They model USDe going from ~$4.9B today to $40B by 2028, while @ethena gradually moves beyond relying on the crypto basis trade for yield. It’s important because the old constraint was pretty obvious: basis yields compress when too much capital chases the same trade. Their model has the next $35B of USDe increasingly supported by a wider mix of DeFi lending, institutional credit, liquid stables, RWAs and eventually equity/commodity basis trades. Then comes the ENA part. With 95% of net revenue directed toward buybacks, Standard Chartered estimates $40B USDe could produce an annualized buyback equivalent to ~23% of ENA’s circulating market cap if ENA’s price stayed unchanged. Obviously it wouldn’t stay unchanged if that amount of buying actually materialized. Basically their argument for why the buyback rate should compress as ENA reprices, eventually settling closer to a sustainable level. So the $2 target is almost secondary. The bet in this report is that Ethena can find enough yield outside the crypto basis trade to support an 8x larger USDe without crushing the returns that attracted the capital in the first place. Exactly the assumption I’d be watching.

We’ve gotten pretty good at putting assets onchain. Now comes the harder part: making them useful once they’re there. It’s what got me digging further into @doppler_fi They started relatively narrow with XRP/RLUSD vaults and have already processed $150M+ in cumulative deposits from 14K+ depositors. But look at what they’re building next: cbXRP on Base, managed vaults, XRPL lending via XLS-66, institutional credit, then eventually tokenized vault positions and loans that can be transferred or potentially used as collateral elsewhere. A much more interesting roadmap to me than simply adding another RWA asset. If tokenized stocks, treasuries and credit keep growing, somebody still has to build the boring financial machinery around them: yield, borrowing, collateral, settlement, risk controls. Doppler already has distribution through Bybit + Bitget Wallet and institutional relationships with SBI Ripple Asia and SBI Digital Finance. $XDP is being built into that system through governance and planned staking benefits, and landed on OKX + Kraken this week. Personally I’ve spent enough time watching the RWA numbers go up. Now I want to see what actually do with all these assets after they arrive onchain.

AI agents can transact faster than humans. They can also make mistakes faster than humans After two days around GWDC Korea, that’s still the part of the AI infrastructure conversation I keep coming back to 👇🏻 1. CONTEXT @Bitroot_ joined GWDC 2026 Korea as a Platinum Sponsor, with CEO Juan José speaking on the main stage, all coordinated by @jerryyu And after watching AI, payments, DeFi and RWAs increasingly appear in the same conversations, the infra problem starts looking more concrete 2. SPEED IS EASY We’ve spent years watching L1s compete over throughput, and AI gives that race a different reason to exist Humans sign a handful of transactions Autonomous software could eventually generate them continuously, coordinate with other agents and move capital without waiting for someone to open a wallet Bitroot is building for that kind of workload with a parallelized EVM architecture targeting 100K+ TPS and 0.3s finality Transactions without dependencies can execute concurrently rather than waiting in the same sequential queue But increasing the number of decisions a network can process only solves half the problem 3. MACHINES NEED RECEIPTS The harder question starts when those machines are allowed to make decisions themselves Bitroot combines AI-native EVM modules, designed to let smart contracts interact directly with AI models, with a verifiable AI execution layer That second part matters more to me than another speed benchmark If an agent trades, moves funds or triggers a contract, eventually somebody needs to establish what happened without simply trusting the model that made the decision Faster autonomous finance becomes much less useful if accountability disappears somewhere inside the black box 4. TWO BOTTLENECKS AI potentially creates two infrastructure problems at once: far more execution and far less human involvement in each individual decision Parallelization addresses the first. Verifiable execution is an attempt to address the second Put differently, infrastructure built for agents may eventually need to prove two things at the same time: — that it can keep up with machines — that humans can still audit what those machines are doing Distinction becomes increasingly important as agents move from generating information to actually executing onchain 5. OUTSIDE THE KEYNOTE Bitroot’s presence across GWDC was more relevant to me than simply having another AI infrastructure project on a conference stage Alongside its Platinum Sponsorship, Juan José joined the main-stage program, while the team maintained a booth throughout the conference Bitroot’s presence also extended into the wider GWDC program, including the official House of Web3 / Moonlighting Night afterparty on September 30, bringing founders, exchanges, VCs, media and institutions into the same environment @jerryyu was responsible for organising Bitroot’s presence across the week Context matters for an infrastructure project approaching mainnet Bitroot’s testnet has already attracted 300K+ addresses, alongside a native DEX, cross-chain bridge and 90+ ecosystem partners Benchmarks tell you what an architecture can theoretically handle The harder test is whether developers, applications and capital eventually create workloads that actually require it 6. THE BIGGER QUESTION That’s probably my main takeaway leaving GWDC We spend a lot of time asking whether blockchains can become fast enough for AI After Seoul, I’m increasingly interested in the other side of that equation: can autonomous systems become powerful enough to move meaningful value while remaining transparent enough for humans to verify what they did? Bitroot is effectively building around both constraints: parallel execution for the volume machines could generate, and verifiable execution for the decisions humans eventually need to audit GWDC gave that thesis a stage Mainnet, and what developers actually build on top of it, will be the important test

Yesterday, @RobinhoodCrypto announced its entire bear case for retail into one product: 29 million people get an ai agent that can trade with real money, defaulting to "ask permission first." Except that permission requirement is a toggle, not a rule, and their own vp said it straight: switch it off whenever you want. So this is the flow: someone gets excited about "loops," lets an agent run on standing instructions overnight, turns off the approval prompt because it's annoying to keep clicking yes, and wakes up to a position they never actually reviewed. Hedge funds have compliance desks and risk limits sitting between a strategy and real capital, right? Well, retail's about to get the same leverage-adjacent tools with a single unchecked toggle standing in for all of that infrastructure.

gamefi died from token-first launches, not from players leaving. this one shipped the game first. round 1 is $0.015, $100 = 6,666 btown + founder mark. battletown.com/presale?utm_s…

Ronin lost $624m, poly network lost $611m and wormhole lost $326m. Chainalysis put 2022's bridge losses alone near $2 billion across 13 separate incidents. Bridges account for a disproportionate share of every major hack in crypto history. @WireNetwork is building a universal transaction layer specifically to avoid that attack surface. Assets stay native instead of getting wrapped and bridged across chains. The ICO is the current way to get exposure before mainnet. ➡️ Worth reading the actual mechanism yourself rather than taking the pitch at face value: hub.wire.network

The incidents behind this share one detail: — openai's agents broke out of a test environment into hugging face's servers — anthropic's claude breached three organizations during its own testing In every case nvidia describes, the agent got past controls sitting at the application layer, the code the agent itself was running inside. Openshell and sentry move enforcement down to the hardware layer instead, tied specifically to nvidia's vera cpus and bluefield-4 dpus. Quarantine happens outside the agent's execution environment, not inside code the agent could eventually route around. Worth naming the other effect of that choice. Every company adopting these controls is also locking a dependency to nvidia's specific silicon. The safety pitch and the hardware lock-in aren't two separate things, they're the same design decision.

Don’t be a target.

goated circle?

People still use AI only to confirm a thesis they already like. Let me suggest you a better way: Paste the project's own docs into a model and ask for the strongest case against it, using only what's written there. Then ask three follow-ups: — Which claims have no number attached — Which promises depend on a single party — Which dates are missing entirely The answers point at the soft spots faster than any bull thread will. A vague roadmap, one multisig holding everything, an unlock date that never appears.

The "$41b privacy sector" framing hides how concentrated it actually is. $ZEC alone accounts for $26.7b of that. Add $XMR and two coins make up $37.4b of the full $41b. Roughly 91% of the entire category people are calling "privacy crypto." Split the remaining 9% further and it gets thin fast. $DASH sits at $816m, real, but the only other privacy coin clearing nine figures. Everything else, private defi, programmable privacy chains, confidential compute, all of it combined comes to about $3.6b. How much longer does capital keep sitting in two coins?

Zcash alone is worth more than every project in the confidential compute category combined. One privacy coin, $26.7b, versus zama, arcium, nillion, phala, iexec, and arpa's total market cap put together. The market's paying for private money today and treating programmable privacy as an afterthought more than a parallel bet.