
founder @ShelbyStudioHQ FC Barcelona
Joined September 2024
- Tweets9.6K
- Following434
- Followers975
- Likes40.5K
β π―ππππ_π―πππππ βͺ retweeted
Compute moves. Jobs get rescheduled. Capacity shifts.
Can your data keep up?
Talk to us β
Imagine your employer publishing everyone's salary on a public website every payday.
Your salary.
Your manager's salary.
Every contractor.
Every payment the business makes.
Most companies would never consider operating that way.
Yet fully transparent blockchains can expose similar financial information when businesses transact onchain. That's a serious problem for adoption.
Companies need confidentiality to manage payroll, suppliers, treasury operations, and sensitive financial relationships. That's what institutional privacy is for.
Enter: Confidential Assets, a standard designed for encrypted balances and transfer amounts.
A zero-knowledge proof confirms each transfer is valid, while the amounts stay encrypted and sender and recipient addresses stay public. It's opt-in, and it's live today with APT.
It's designed for compliant use cases like payroll, treasury, and B2B settlement.
This is about more than encrypting numbers on a block explorer.
It's about bringing the confidentiality companies already expect from financial infrastructure onchain.
β π―ππππ_π―πππππ βͺ retweeted
I need to go out more, speak to people more, laugh more and call my friends and family more
β π―ππππ_π―πππππ βͺ retweeted
Okay I got in π€
Fungo Points are live π¬
β Unlimited round prizes: every mint is a ticket, every round pays ETH
β Points level you up and open boxes
β Top 3,000 share 1B FUNGO
β Genesis holders: $1,000 box every week
Full breakdown in our founder @0xFlydev 's article below π
β π―ππππ_π―πππππ βͺ retweeted
Grateful to mutuals that always show me way π.
Thank you so much @AwakDaniel_
β π―ππππ_π―πππππ βͺ retweeted
Before an AI touches your evidence, can you prove where it ran and who could read it? Today you canβt. I built Custos Seal at the NEAR HACKATHON to fix that. @NEARProtocol #NEARHackathon π§΅
Who remembers this guy? All of you in the comment section were encouraging his stupidity back then
How e con be now? π
β π―ππππ_π―πππππ βͺ retweeted
Compute may be the most important constraint in the agent era. But data is increasingly the constraint on compute.
Batching, off-peak hours, cheaper GPU-hours wherever they are: these compute optimizations depend on workloads being able to run where capacity is available. If every change in compute means another cycle of moving, copying and staging data, those efficiencies get harder to realize.
Compute is dynamic. Data needs to keep up.
If you're building in AI, the most important thing to master is the economics of compute. My take: respect the GPU or the GPU will humble you.
We're in a compute shortage. A GB300 costs roughly $7/hour on the spot market so depending on the frontier model, that hour buys you some millions of output tokens.
The only question is: how much customer value can you create with those tokens?
If it's less than $7, you have a problem. Customer won't pay you, so you won't be able to afford more compute, and VCs can subsidize you forever. Eventually, someone who creates more value per GPU-hour will outbid you.
The same GPU-hour could be spent generating another meme or helping discover a cancer treatment. The economic value of those outputs is wildly different.
My take is that you have to design your entire product around this equation. What's the smallest model that can do the job? Can you batch the work? Run it asynchronously during off-peak hours? Avoid generating tokens you don't need?
Every architecture decision, every feature, every token changes your compute economics. The entire roadmap has to be decided for the compute equation.
What's hard is that in traditional software eng, you could often treat compute as practically free but the agent era, compute is the most important constraint.
β π―ππππ_π―πππππ βͺ retweeted
For high throughput storage systems, there is an overarching goal -- move data at line rate (i.e. your network capacity). It is easier with NVMe caches, but doing that with HDDs is where you get a massive price + performance advantage. Doing this globally, is an entirely different challenge.
Shelby demonstrates you can saturage a gateway and then horizontally scale. There is no system throughput bottleneck per se..
AI labs, neoclouds, token factories - they all move PBs of data/day across DCs. Model weights/checkpoints, customer data, outputs..
Single address, global access, performance boudless..
β π―ππππ_π―πππππ βͺ retweeted
We ran out of network before we ran out of Shelby.
One gateway. 128 concurrent reads. ~2 TiB of data.
~46 Gbps of payload throughput across 2Γ25 Gbps of client-facing network. Client-link traffic reached 47β49 Gbps.
The network hit its limit first.
Full performance breakdown β
β π―ππππ_π―πππππ βͺ retweeted
Enterprise AI at @teepin. Distributed 3D rendering at @pictor_network. Spatial intelligence at @PathPulse_AI.
Three @shelbyserves Private Beta workloads, coordinated on @Aptos. One common need: data that stays usable as compute changes.
Inside each one: shelby.xyz/news/shelby-privaβ¦






















