Do You Want To Get Lost?

i’m earning 33.88% on my stables on @aeroxyz please don’t dilute me

onedollaraudit.com is now running on @pashov solidity auditor v4 pay 1 usdc and get a state-of-the-art first-pass ai audit put this in your deployment pipeline and thank me later you can even stake.onedollaraudit.com and get it free

Alchemix v3 is live on @base! made a little explainer why it is fucking based. sound on 🧙‍♂️ ⚗️

Gearbox is a great example of why you shouldn’t confuse lack of attention with lack of progress, the market hasn’t exactly treated them kindly, but while everyone moved onto the next narrative they just kept grafting on the same core architecture. This RWA leverage solution they’ve just released is where that strategy is really going to pay off imo. I’ve spent a bunch of time looking into the different approaches being built here but what @GearboxProtocol has done is by far the most elegant design I’ve seen. It’s not just better looping ux, looping naturally assumes the underlying asset is liquid and everything settles atomically so you borrow, buy the asset, deposit it again and repeat, which is great for crypto native assets but doesn’t really work for RWAs where you have different redemption timelines, transfer restrictions, KYC requirements and in a lot of cases barely any secondary liquidity. So gearbox gets around all these issues by using the credit account to borrow the full amount upfront and subscribe directly with the issuer. Instead of needing 5 or 10 or however many separate loops and waiting through the settlement process each time, you can create the entire leveraged position in one go and redeem the entire thing in one go on the way out. What I really like about this is that leverage no longer needs to depend on secondary market liquidity to the same extent, because it was always pretty silly to expect an issuer to bootstrap $50m or $100m of dex liquidity before people can take levered positions on the asset. With this model the position can scale against available credit instead because the credit account is interacting directly with the issuer, which means the lending side can support assets that would have been impossible or just really inefficient to support through the normal loop model. Even the asset specific stuff like KYC, transfer restrictions, redemption and specialised liquidation logic can sit inside gearbox’s infra rather than the lending protocol having to figure all of that out. And because gearbox is separating the asset specific execution and risk machinery from the funding layer, I think the bottleneck for lending protocols changes quite a bit too. Because the problem is no longer whether an asset has enough secondary liquidity to be listed and what really starts to matter is whether somebody can actually originate good borrow demand against the liquidity sitting there. If an issuer can bring the asset, gearbox can handle the market specific plumbing and a lending protocol can provide the funding, which is just a more efficient way of creating leverage around assets that never really fit the normal money market model in the first place. These guys have done a great job of addressing the actual market specific pain points and creating a structure where everybody wins, RWA issuers get leveraged distribution without needing deep secondary markets first, stablecoins get a new structural source of borrow demand and lending protocols get access to a much wider set of credit opportunities. What I’m really excited to see now is how much leverage this can support in practice, because if you can get a decent amount of leverage on assets that were basically unleveragable before then you open up a pretty massive new market. Really excited to see how this plays out and wishing the chads at gearbox all the best with this launch.

MLX-Serve 26.9.5 is out, and this one is a pretty serious update for anyone running local AI on Apple Silicon. The Desktop edition now adds support for Qwen-Image-2.1, so image generation is becoming part of the same local serving stack instead of needing a completely separate setup. It also ships with Prism Bonsai 2 support. That is the 27B Qwen3.8-based model compressed down to roughly 6GB, which is already interesting for local inference. The MLX implementation is being positioned as one of the fastest ways to run it, although the actual speed will depend on the Mac, context size and workload. The other change I’m watching more closely is concurrent inference. MLX-Serve has been pushing beyond the “one request at a time” local-server experience, and this release improves how multiple streams are handled. That matters a lot if you’re using a Mac as an actual AI server rather than just chatting with a model. And finally, there are now charts and evaluation results. That’s a welcome addition. Raw tok/s numbers are useful, but seeing the benchmark methodology, scaling behaviour and model comparisons makes it much easier to understand what the runtime is actually doing. So this release is less about adding one flashy model and more about turning MLX-Serve into a broader local AI stack: Qwen-Image-2.1 for vision generation. Bonsai 2 for lightweight 27B inference. Better concurrent streams for serving multiple requests. Charts and evals to show the performance instead of just claiming it. If you’re running Apple Silicon and experimenting with local models, MLX-Serve is definitely worth watching. And yes, apparently the GitHub link is the kind you should click before it mysteriously disappears.

I alluded to big things happening this week, and it is time to tell you what. @AlchemixFi is launching on @base later this week. I'm particularly excited about this deployment, which I have detailed why in the article posted in the 2nd tweet of this 🧵 👇 1/11

Created this little generative village for SANROK's new web. Progress is going smooth ⌨️

lmao 2 years of building in stealth and your moat lasts 2 days

Weekly Recap @RenzoAI 1. Expanded basis trade on @HyperliquidX for $ZEC and $ETH that are paying +18% APR 2. Unified accounts were enabled allowing automated rebalancing across multiple assets per risks guards User growth and adoption continue to increase with a lot of interest in additional assets, @Lighter_xyz and a curated vault. We've been running a proof of concept with a few million dollars that is ideal for the current state of the markets where every other token is pumping. Might need to bring that to market sooner than we had anticipated Appreciate all the positive feedback...hgher 🤌🏻

Arbitrage and sandwich attacks are both categorized as MEV, but they have fundamentally different effects: one improves market efficiency, while the other exploits user transactions. Arbitrage responds to price discrepancies that already exist. It brings prices across markets back into alignment. A sandwich attack creates an artificial price movement around another user’s transaction, forcing them into a worse execution. Bancor’s infrastructure treats them accordingly. @CarbonDeFixyz's one-directional orders are immune to sandwich attacks by design. The Arb Fast Lane searches for arbitrage opportunities across Carbon DeFi and all major DEXs on a network to help fill orders, keep strategies actively trading, and maintain price alignment across the chain. One manufactures market inefficiencies. The other corrects them.

JUST IN: 🇺🇸 CFTC permits developers to build passive derivatives software without registering as brokers, including for crypto markets.

If you think this new shirt design is cool - please share (lisa needs braces). This is a rework of an early Encarta logo to be Vitalic - riffing on an old Based Money meme, in the style of DYWTFS. I'm trying to enable international purchases and this is as cheap as I could make the shirt. sgt-slaughtermelon.myshopify…

Union Alpha (stealth model) is free for the next week - no data training - built for agentic coding - supports images let's see what you can do

A warm welcome to all the new Renzo community members! Here are a few facts: 1. @RenzoAI raised +$20m from @yzilabs, @galaxyhq, @Maven11Capital, @OKX_Ventures and a handful of other tier 1 VCs. 2. $REZ launched in April 2024 and all investor unlocks are well past us, with the team still vesting. 94% of $REZ is in circulation. 3. Exchange listings include @binance, @coinbase, @Official_Upbit and @BithumbOfficial, with more in the pipeline. 4. The team is fully doxed, check X affiliate accounts, and mostly US based OG builders. 5. We originally launched $ezETH, which grew to $4b in TVL. Now we’re building a self-custody onchain basis trade product on @HyperliquidX, @Lighter_xyz and @RobinhoodApp aka decentralized $ENA. Our DMs are always open. We look forward to growing the community and building products you love 🤌🏻

Weekly Recap @RenzoAI 1. Announced RenzoFinance rebrand 2. Launched MVP of basis trade on @HyperliquidX for $BTC and $HYPE 3. @Official_Upbit ethereum:0x3b50805453023a91a8bf641e279401a0b23fa6f9 USDT listing Overall, a very positive week. Initial numbers for users and volume are ahead of our internal September goal. The product is still very much an MVP, but a few large upgrades are in the pipeline this month that I’m excited about. The most positive note is that a couple of neobanks and LPs reached out looking for new products to offer their users. We had a few issues with the new URL being flagged as malicious due to the previous domain owner, but that should now be fully resolved. We’re also investing in education, as most users don’t understand what a basis trade is or where the funding rate comes from. DMs are open for everyone… higher 🤌🏻

concerning. this world computer ethereum thing sounds very dangerous and left to its own devices could disrupt all of global finance and the economy writ large

im working with a municipality on a new anti-wildfire initiative, they want to raise $5m-$10m in bonds on blockchain for the project. id like to line them up with a partner thats built on ethereum. who should i send them to?

Next upgrade will push APY's to +10% via portfolio margin app.renzofinance.com/