Drizzle

[Sui @SuiNetwork]

Sui public mainnet just reached a peak of 6M+ TPS.

Watch the livestream to see.👇

[Sui Developers @suidevelopers]

Reminder for builders ⚠

JSON-RPC will be disabled on Sui Foundation full nodes on Testnet, including public-good https://t.co/ZTD66Dndie, at 12 PM EST on July 8.

If your app or workflow still depends on JSON-RPC, migrate to gRPC, GraphQL RPC, or the Archival Service now.

Migration guides:
https://t.co/3qhucNr7d8
https://t.co/qS6swasr1Q

[Walrus 🦭/acc @WalrusProtocol]

Your agent still forgets everything the moment a session ends.

Walrus Session 5 is still open!

Write one prompt that fixes a real problem, and show it works.

$2,000 WAL pool. Submit by July 13.

Details 👇

[dealbreakr @emeieiron]

To understand MEMORA, you first have to understand why AI memory is broken

Today's systems struggle to maintain a coherent relationship with users over time because they sit between two extremes: specificity and abstraction

Standard RAG over-indexes on specificity, storing isolated facts. They are accurate, but stripped of context and difficult to navigate. Summarization does the opposite, compressing everything into a generic profile that is too abstract to be useful

This is the representational gap. AI either drowns in disconnected details or relies on oversimplified summaries

MEMORA addresses this with a harmonic architecture that keeps memories rich and intact while building a separate structural layer dedicated to navigation, linking high-level concepts with low-level details

[DeepBook Protocol on Sui @DeepBookonSui]

Sui settles in about 400 milliseconds.

See what that speed means for trading on the DeepBook App, and join the waitlist for early access: https://t.co/ZtquuCmDOv

[dealbreakr @emeieiron]

I hate to choose between elegance and efficiency with tooling

[Sele @iv_dev3]

.@taliseio send ux.

check out the TestFlight app: https://t.co/W2C2E0jzx6

[Turbos Finance @Turbos_finance]

Introducing Stable Range Farming on Turbos!

Stable pool LPs know the pain: a minor depeg knocks your position out of range, cutting off rewards — until you rebalance. Again.

Stable Range Farming ends that. Set it once. Let it farm.

🧵👇

[Adeniyi.sui @EmanAbio]

What if your AI answered to you, not a moderation team? Meet @audricai . Private, uncensored, fully yours. Every model worth using, open and closed. All under one roof.

[Sui @SuiNetwork]

Business moves deserve privacy.

If everyone can track your transfers, you lose your edge.

Confidential transfers on Sui. Live in public beta.

[evan.sui @EvanWeb3]

Excited to share Hashi has received more support from key institutional partners @CumberlandSays, @swissborg, and @0xfluid

[DeepBook Protocol on Sui @DeepBookonSui]

One of the first markets on the DeepBook App: will Bitcoin go up or down?

You make that call directly on DeepBook’s order book, which has powered $17B+ in Sui trades.

[Sui Developers @suidevelopers]

Sui Overflow 2026 https://t.co/G3x6Bu55QK

[Sui Developers @suidevelopers]

Confidential transfers on Sui just hit public beta on Devnet.

Open source code, docs, and an example app are live now.

What you need to know to start building: 🧵

[Kostas Kryptos @kostascrypto]

Interestingly @WalrusProtocol can open the door for future-audited and verifiable lightning channels. One can use it to periodically store offchain state, then verify it on Sui or other chains.

[Kostas Kryptos @kostascrypto]

I’ve see many discussions last week that formal verification is the only thing needed… but hear me out:

Formal verification is extremely valuable, but not a panacea.

have seen this first-hand in my Silicon Valley career, including at Facebook, where some systems had extremely serious engineering talent behind them and, in some cases, heavy formal verification efforts. And still, things could be missed. Not because people were careless, but quite the opposite. Some of the best engineers and researchers in the world were involved.

The issue is deeper: someone still has to define the rules: formal methods can prove that software satisfies a specification.

They CANNOT magically prove that the specification captured every real-world requirement, every adversarial behavior, every edge case, every dependency, every environmental assumption, and every way humans will compose the system with other systems.

Note that many vulnerabilities do not necessarily come from code failing to follow the spec. They come from the spec being incomplete.

You can prove that a smart contract preserves an accounting invariant, but miss that the oracle price can be manipulated.

You can prove that a protocol transition is valid, but miss that the bridge dependency can be compromised.

You can prove that a cryptographic implementation matches an algorithm, but miss that the randomness source is biased or reused.

You can prove memory safety, but miss a side-channel leak.

You can prove that a wallet signs only valid transactions, but miss that the UI tricks users into approving the wrong intent.

You can prove that a consensus rule is implemented correctly, but miss an economic attack around incentives, MEV, or validator behavior.

You can prove that your code is correct, but still depend on an unverified compiler, library, VM, OS, hardware wallet, cloud service, API, or package maintainer.

Nobody modeled that weird input.
Nobody considered that timing assumption.
Nobody captured that economic incentive.
Nobody thought about that upgrade path.
Nobody formalized that interaction with an external component.
Nobody audited the entire dependency tree.
Nobody asked whether an upstream dependency could be compromised and silently invalidate the whole security story.

So yes, formally verified software can absolutely still have vulnerabilities. In short, “formally verified” does not mean “secure”

It means: secure with respect to the properties someone remembered, understood, and precisely defined, under the assumptions they made about everything around it.

Pease please… do not replace threat modeling, adversarial review, dependency auditing, supply-chain security, fuzzing, continuous audits, testing, economic analysis, and engineering humility with a certificate that says “ah it’s verified”, bulls**t

The hardest part is often not proving that the code is correct. It is defining what “correct” truly means and what the code is allowed to trust.

What you need is a top notch engineering and red hat teams, whoever doesn’t have them at L1, bridge, or wallet layer they are mathematically dead ☠️. And I highlight that finding the top talents in cybersecurity is hard and expensive ;)

[evan.sui @EvanWeb3]

We are not slowing down. Today, confidential transfer is going live on Sui Devnet.

Sui is not your grandpa's blockchain. It's a coordination layer, settlement layer, a programmable platform where all the tools you need to create/exchange/manage assets of value. Confidential transfers is the latest building block added to the stack. A lot more to come.

[Adeniyi.sui @EmanAbio]

if you want to move any major stablecoin today, sui is by far the best network to do that

and THERE IS NO SECOND BEST

[Walrus 🦭/acc @WalrusProtocol]

This is what cross-platform memory under the user's control looks like in real life.

@Conso_xyz pulls activity from social, gaming, AI tools, and onchain sources, runs it through Walrus where it gets encrypted, and lets any agent recall it later. 🦭

[Adeniyi.sui @EmanAbio]

Confidential transfers are coming to #Sui.
The hard part of private money isn't hiding the amount but guaranteeing nobody can mint value out of thin air while the supply is shielded.
We solved it by scoping the cryptography down to one thing: range proofs on transfer amounts. Conservation of supply lives in the protocol, not the proof. Unauthorized mints are impossible by construction.