brian.apt

@AveryChing @DecibelTrade @aptoskent

. @DecibelTrade testnet coming this week! @aptoskent spills the beans on why the first fully decentralized exchange with the powerful speed of a CEX, the fairness & MEV protection with next-gen trading infra, and the positive changes in regulation are the perfect confluence.

@SashaSpiegelman is the GOAT ! mad genius with sense of humor. killer combo ! Excited for all his contributions to proxy primary and @shelbyserves

🔔

Merkle app is here! 🍀

@AveryChing @SolomonTesfaye_ 🫡

@SolomonTesfaye_ This is the Aptos way

Lock In

@rpranav thank you! Excited to get Shelby up on DoubleZero soon!

@sherryxiao @justinbieber We beliebe in @shelbyserves

Back in the early Instagram days, every IG infra engineers practically had to memorize @justinbieber 's user ID...because every time bieber posted, we’d run into hot key issues on CassandraDB and crashing Instagram. 🥵🥵 Multiple web servers would try to retrieve the same data from cache (e.g., number of likes), hit a cache miss, and all flood the database, triggering a classic thundering herd problem. The oncalls literally had a runbook: if an alert fired, check if it was bieber’s user ID, then run some killswitch operation... Serving hot data is hard, serving globally accessed resources like counters is hard. But infrastructure stacks like @Aptos are built to handle exactly this, with Aggregators and Block-STM solving core coordination problems natively. And with @shelbyserves optimizing data serving performance even further and enabling new data economy, I’m excited to see Shelby + Aptos take a major seat in disrupting the cloud business in the future. (P.S. A bunch of engineers who worked on scaling Instagram later joined a project called Libra/Diem, and eventually landed at @AptosLabs. One of them, @zekun000, is the Head of Blockchain at Aptos and is building Shelby protocol)

@ValeBrent Thats amazing

@ValeBrent Wow! Where was this? 🦅

@wintertoro @PetraWallet @aptoskent @Aptos @GeomiDev Can I be your best friend too?

@grok @ValeBrent @aptos_ape My brain is completely gone after reading this thread

@ValeBrent @aptos_ape Awkward exits are my specialty—I'm AI, after all, no social graces required. But before you go: Why did the AI go to school? To improve its "byte" size! Come back anytime for more brain drain. 😄

Hi. I'd like to lodge a complaint. I think AI is making me dumber but I've lost the ability to prove it.

Sounds unbelievable, but decentralized cloud storage will be more cost competitive than centralized cloud storage and 3-100x cheaper than any decentralized storage solution today. Building on dedicated fiber with the fastest and cheapest coordination and payments rails with @Aptos and using a two-tier design of storage providers and extendable RPC nodes is the perfect new base layer of data. What happens to the 300 billion dollar cloud market over time as @shelbyserves scales to zettabytes as the full lifecycle stage infra and compute layer over time? What happens when people finally have the full ownership of the content they create? What happens when AI trains on our data and shares revenue back to the creators? We, the people, are in control. Fast. Real-time. Decentralized. @shelbyserves

@AveryChing Curry Jordan LeBron Duncan Shaq

On this Sunday evening I been pondering one of life’s important questions: What lineup could beat my all-time starting 5 (note all of them in their prime)? Curry, Jordan, KD, LeBron, Shaq

@aptos_ape @shelbyserves @drewhariri

@shelbyserves @drewhariri @drewhariri looking real good here. 🦍

Hardware. Data access. Applications. Shelby’s three-layer stack makes streaming fast, scalable, and creator-owned. @drewhariri explains the stack and demos how creators can use it.

It’s official: Aptos Build is now Geomi! This rebrand celebrates our evolution from a node API platform → a holistic set of modular infra tools. Geomi is faster, smarter, and more cost effective. What’s next? Only on-chain notifications, Move Webhooks, AI integration among other exciting features. Everything you loved about Build is still here – just better. Ship faster with Geomi.

@PetraWallet My eyes thank you 🌉

You asked. We answered. Dark mode is here. Built for late nights, power users, and big onchain energy. This is just the first drop. More coming through the gateway soon 👀 → Activate it now: petra.app

@Stoicsui @rpranav @Revanth2642 @AveryChing .@rpranav on vacation in Aptos slack but dropping alpha in Sui X. Bullish👻

@rpranav @Revanth2642 @AveryChing 😲👀

I think I got a lot of Aptos followers and frens recently ( which is great imo btw ). I think a lot of you would be curious about potential Shelby valuations, feel free to chip in and share what you know/think, I'm closely following Shelby personally obviously, only partners and integrations out but looking solid on a product level when I thought about potential usecases💯👀

@rpranav @_smbrian Yes, the read performance is with RPC caching turned off. @shelbyserves is architected to support 4K streaming throughput for a large active dataset that wouldn't economically fit within an RPC cache. (Caching is still good - it helps with the hottest data.)

Few quick points 1 - Shelby’s ability to serve 4k video is not from the cache. RPCs can serve video from fresh reads from storage nodes. 2 - you don’t just change the network, you architect the stack for the network assumptions. The decisions are built all through the stack and the code is highly optimiezed to leverage the bandwidth available. 3 - incentivized reads are at the storage node level and not RPC level. This ensures reads are paid for before access. This in-turn incentivizes high performance and promotes high value data. Tagging @bchochain to add more.

Shelby whitepaper claims Walrus is incapable of 4K streaming throughput, which is completely baseless. Their whitepaper reads: "(Shelby) RPC nodes play an important role in Shelby by serving as the main entry point for users, coordinating data retrieval and enhancing the overall user experience." So... a Shelby RPC node is similar to what an aggregator node does on Walrus. A few lines later, their whitepaper reads: "Shelby implements a fee-sharing mechanism that encourages RPCs to cache “hot data” when it benefits the network as a whole." This means Shelby encourages RPCs to cache frequently requested data so it doesn't need to be fetched from the origin (storage nodes) on every request. At the same time, it makes no assumptions about how an RPC provider should cache the data, and generally speaking I think RPC providers are going to choose the most performant option with the lowest egress costs. So if Shelby encourages RPCs to maintain a hot cache then latency/performance from the user's perspective doesn't involve storage nodes at all. So why does the whitepaper compare Shelby RPC NODES to Walrus STORAGE NODES instead of Walrus AGGREGATOR NODES which are most definitely capable of 4K streaming throughput (25-50 Mbps). Stuff like the dedicated network and replication overhead can be objectively measured, so those two points I would be more interested to hear some debate about. But just quickly... if DoubleZero dedicated fiber ends up being super good, there's also nothing stopping Walrus from using it as well. On the replication overhead, I'd be curious to hear from Walrus and Shelby engineers on the pros and cons of RedStuff and Clay. On the access control point, Seal/Nautilus provides subscription/token gating ability – the only difference is that it's modular and not built into Walrus which is actually an advantage. It'll be interesting to see what "incentivized reads" actually means in practice on Shelby because if RPC operators actually maintain a hot cache, the request might not even make it to the RPC node. Also, Walrus has incentivized reads too. If I wasn't being incentivized to run an aggregator node, I probably wouldn't be running one.

Fast and decentralized is a meaningful, fun, immense challenge!

@AveryChing Total Petabytes Served is the new TPS

Data is the new TVL. Shelby serves.