We're witnessing the buildup of the largest repository of private data in history, controlled by a few entities.
The world's largest data honeypot, growing daily.
Blackthorn is the substrate making AI inference private by default, with the smallest trust assumptions.
🚨 It's almost time. 🚨
The ZINC Gacha Machine opens soon!
The community told us they like Pokemon cards, so we are bringing them to Zootbox.
Final preps ongoing. Who's ready?
Introducing the Parimutuel Market Maker (PMM).
A new trading mechanism designed specifically for permissionless prediction markets.
Only possible on Solana.
Staked ARX is locked for at least 2 days, unstaking triggers a cooldown with no rewards, and stake can be slashed if your operator misbehaves.
Only delegate to operators you trust.
Get started: https://t.co/1BJNqc71dZ
How to stake:
1. Choose a Node Operator to delegate to.
2. Select the amount of ARX you want to stake.
3. Earn SOL based on the work completed by your chosen Node Operator.
Well-known operators include H2O Nodes, Staking Facilities, and Greenfield (ranked by stake).
The Privacy Show returns!
This week, @catmcgee hosts @milianstx, @herz0g, and @typi_cal_ to talk all things privacy on Solana
Today at 2PM ET here on X
Token holders can now delegate stake to nodes they trust, earning a pro-rata share of computation fees (net of operator commission).
Note: Delegated stake is subject to slashing and may be partially or fully lost if a node misbehaves or fails protocol requirements. Delegation is not a deposit and is not principal-protected.
Delphi has released an in-depth report on Arcium, covering our tech stack, confidential payments, ecosystem, and Blackthorn, our newly announced encrypted AI inference architecture.
The full report is free to read on the Delphi portal linked below.
Arcium is the encrypted compute layer for blockchains.
Everything on a public blockchain is permanently visible. That works for speculation but rules out anything that has to keep a secret while it runs. A sealed-bid auction collapses once the bids are visible. An order gets front-run once its size is public. None of this can move onchain because the transaction exposes the secret the moment it is visible.
@Arcium separates verification from execution. The outcome stays public and verifiable, while the logic that has to stay hidden runs privately across a cluster of nodes where no single node reconstructs the full data. Arcium's live backend (Cerberus) keeps the result correct as long as one honest node remains. Settlement happens onchain, but the inputs stay private.
This opens up a new design space for public blockchains. A launchpad can launch as a sealed-bid auction where no bidder sees the other bids until it closes. A DEX can process an order privately and reduce the surface for MEV. A company can run payroll onchain without exposing its pay structure.
Privacy-oriented DeFi has tried to address this by walling off everything, but this came at the cost of composability and fragmented liquidity. Arcium keeps settlement public and composable while the sensitive logic stays hidden. It lets a public chain keep a secret.
1,500,000 Encrypted Computations on @Arcium.
Another massive milestone for Zinc, achieved just 6 weeks after launch.
Arcium is a core part of the Zinc infrastructure, and both products continue to push each other forward every day.
🦺☂️
Removing the CPU TEE takes out one of the biggest points of failure in Private AI today, making it both more private and more available.
Blackthorn is coming soon: https://t.co/LuIvv8Unxt
Every confidential AI cloud today trusts a CPU TEE somewhere in the stack.
Those chips have had over a dozen published exploits since 2020, including key extraction and forged attestation.
Arcium Blackthorn removes the CPU TEE from the trust chain entirely.
Because the CPU TEE is no longer required, Blackthorn is built to run on NVIDIA GPUs already deployed today.
Right now only a small fraction of GPUs qualify as confidential-AI-ready, since the CPU TEE and its certified cloud stack narrow the pool sharply.
Removing that dependency opens the door to the full multi-million GPU supply for Encrypted AI inference, with privacy guarantees stronger than today's state of the art.