
Founder @Dhest_den || CM || 随时欢迎您的联系 ||
Web3
Joined May 2024
- Tweets16.3K
- Following3.2K
- Followers3.1K
- Likes49.1K
Pinned Tweet
In web3, projects need more than visibility, a clear purpose, strong execution, active communities, and a commitment to delivering real value.
What matters is identifying projects with real intent to build and following their progress beyond the timeline hype.
My attention is on @themutualfun as I explore projects working toward the next chapter of Web3. I'm interested in learning more about its vision and the opportunities.
@themutualfun is building a permissionless, on-chain mutual fund ecosystem designed to bring a new approach to collective participation in DeFi.
Their idea is simple and interesting, they combine NFTs, on-chain finance, and community-driven decision-making in one ecosystem.
This concept is worth watching because it explores how blockchain can make financial participation more transparent and community-driven.
Staying informed, joining meaningful conversations, sharing ideas, and supporting innovation
Numerous privacy tools force institutions into a bad tradeoff: expose your positions, or hide everything and lose the ability to operate.
With real capital, neither works.
Institutions need to move funds and earn yield without revealing their strategies.
A visible position can be tracked, front-run, or copied. A strategy loses its edge the moment competitors can read it live.
They can also operate under oversight.
Auditors need to confirm numbers. Regulators need to inspect specific values when there's a legitimate reason.
Total secrecy isn't an answer to that.
The requirement is two-sided:
Private by default.
Verifiable on demand.
Full transparency kills strategy. Full secrecy kills compliance. You need both to hold at once.
FHE solves the first half.
Balances and amounts stay encrypted on-chain. Smart contracts update them through homomorphic operations, meaning computation happens directly on encrypted data.
The system keeps running. Funds keep moving. Balances keep changing.
But the actual numbers stay hidden by default. An outside observer sees only ciphertext.
What's missing is a controlled way to reveal one specific value later.
One balance for an audit. One amount for a regulator. Narrow, deliberate, and tied to proper authorization.
Think of it as a sealed envelope that can be processed without being opened.
Its contents stay private unless access to a specific value is authorized and recorded.
@primus_labs builds an access-control layer on top of encrypted state.
Decryption stops being the default and becomes an explicit, auditable exception. Authorized, logged, and limited to what was permitted.
The result makes institutions keep strategies confidential while giving auditors and regulators a reliable path to verify what they're entitled to see.
Privacy and accountability working together is what brings real capital on-chain.
Most privacy tools force institutions into a difficult tradeoff: expose their positions, or hide everything and lose the ability to operate.
Neither works with real capital.
Institutions need to move funds and earn yield without revealing strategies, while still allowing auditors or regulators to check specific values when authorized.
FHE keeps balances and amounts encrypted on-chain. Smart contracts update them through homomorphic operations. The actual numbers stay hidden by default.
But encryption alone is not enough. You still need a controlled way to open a specific value later.
Think of it as a sealed envelope that can be processed without being opened. Its contents remain private unless access to a specific value is authorized and recorded.
Primus builds an access-control layer on this encrypted state, so decryption becomes an explicit, auditable exception rather than the default.
primuslabs.xyz
Your Data Should Be Useful Without Being Exposed
One thing increasingly important as digital technology evolves is the amount of personal information people have to reveal before they can access useful services.
Consider an AI application that needs specific details to generate an answer or a financial platform that has to analyse confidential records. Does performing these tasks really require giving the service unrestricted access to the original information?
I believe there's a better way to approach this.
Fully Homomorphic Encryption (FHE) introduces a different model of data processing. Instead of decrypting information before using it, computations can happen directly on encrypted data. This means useful operations can be performed without revealing the underlying information to whoever handles the computation.
That's what makes the work around @primus_labs particularly interesting to me.
Through its open-source Aegis framework, Primus Labs is exploring ways to make software programs and machine learning models compatible with encrypted computation.
The possibilities for AI are worth considering. Imagine interacting with an AI service, receiving meaningful outputs, and keeping your original personal information protected throughout the process.
The same concept could be valuable in finance, where applications need to analyse sensitive information without exposing more details than necessary.
Of course, FHE still faces important limitations. Computational overhead, processing speed, and the range of efficiently supported operations remain significant challenges. It's not a universal solution today, but its potential is difficult to ignore.
As our digital lives become increasingly dependent on data, protecting information should go beyond securing databases and preventing breaches. We also need to rethink whether services should see the information they process in the first place.
Why should accessing the benefits of technology require sacrificing control over our personal data?
This is one reason encrypted computation and the work being done by @primus_labs deserve more attention.
AI agents are gradually moving beyond simple automation and beginning to execute tasks that involve actual financial assets. But as their responsibilities grow, one concern becomes increasingly important: how can we confirm that an agent did exactly what it was instructed to do?
Consider an AI agent handling a DeFi transaction on your behalf. It collects data, connects with external platforms, and executes several operations. The transaction might be successful, but that doesn't automatically prove the agent used reliable information or respected every condition you specified.
When real assets are involved, a successful outcome alone shouldn't be the only measure of trust.
That's one reason @primus_labs and its Veritas approach caught my attention.
The idea of verifiable AI agents is particularly interesting because it focuses on making their execution process independently checkable. Instead of simply accepting an agent's claim that everything was done correctly, users could have evidence showing what actions were performed and whether the required instructions were followed.
This could become especially valuable in crypto environments where agents interact with smart contracts, retrieve external information, and execute transactions under specific rules.
Of course, being able to verify an agent's actions doesn't automatically guarantee that its decisions are correct. The quality of its data and the instructions it receives remain equally important.
However, transparency and accountability can significantly reduce the need to rely on blind confidence.
As autonomous agents take on more financial responsibilities, proving what they did may become just as important as their ability to execute tasks.
Would you feel comfortable allowing an AI agent to manage your crypto transactions if it could provide verifiable proof that it followed your instructions, or would you still want to manually approve every transaction?
AI agents are gradually moving beyond simple automation and beginning to execute tasks that involve actual financial assets. But as their responsibilities grow, one concern becomes increasingly important: how can we confirm that an agent did exactly what it was instructed to do?
Consider an AI agent handling a DeFi transaction on your behalf. It collects data, connects with external platforms, and executes several operations. The transaction might be successful, but that doesn't automatically prove the agent used reliable information or respected every condition you specified.
When real assets are involved, a successful outcome alone shouldn't be the only measure of trust.
That's one reason @primus_labs and its Veritas approach caught my attention.
The idea of verifiable AI agents is particularly interesting because it focuses on making their execution process independently checkable. Instead of simply accepting an agent's claim that everything was done correctly, users could have evidence showing what actions were performed and whether the required instructions were followed.
This could become especially valuable in crypto environments where agents interact with smart contracts, retrieve external information, and execute transactions under specific rules.
Of course, being able to verify an agent's actions doesn't automatically guarantee that its decisions are correct. The quality of its data and the instructions it receives remain equally important.
However, transparency and accountability can significantly reduce the need to rely on blind confidence.
As autonomous agents take on more financial responsibilities, proving what they did may become just as important as their ability to execute tasks.
Would you feel comfortable allowing an AI agent to manage your crypto transactions if it could provide verifiable proof that it followed your instructions, or would you still want to manually approve every transaction?
A part of the flaws I believe Web3 projects need to solve is making innovation accessible without forcing users and developers to abandon the systems they already depend on.
Think about a blockchain application that has spent years building its ecosystem. It has users, liquidity, smart contracts, and an established presence on a particular network. Introducing a new capability shouldn't automatically mean migrating everything to another chain, spending additional resources, or creating a completely different experience for its community.
This is where privacy-focused infrastructure becomes particularly interesting.
@primus_labs's approach to verifying information obtained from web services and enabling computations without unnecessarily revealing sensitive data is rop
Consider a lending platform that needs to evaluate someone's financial background before approving a loan. Traditionally, verifying such information could require users to disclose far more personal details than the application actually needs.
Privacy-preserving verification offers a different possibility: proving that specific requirements have been satisfied while keeping the underlying information confidential.
Even more importantly, these capabilities should be accessible to existing applications without forcing developers to redesign their entire infrastructure or migrate their users to another blockchain.
There are still important challenges to overcome, including security, scalability, integration costs, and performance. However, making privacy technology compatible with the applications people already use is a meaningful step toward broader adoption.
The future of Web3 shouldn't require users to sacrifice convenience or control over their personal information just to benefit from better technology.
That's why I'll be paying attention to what @primus_labs continues to build.
Here's an interesting question: Should privacy become a standard feature that every Web3 application can integrate into its existing infrastructure, rather than something that requires an entirely separate ecosystem?
A part of the flaws I believe Web3 projects need to solve is making innovation accessible without forcing users and developers to abandon the systems they already depend on.
Think about a blockchain application that has spent years building its ecosystem. It has users, liquidity, smart contracts, and an established presence on a particular network. Introducing a new capability shouldn't automatically mean migrating everything to another chain, spending additional resources, or creating a completely different experience for its community.
This is where privacy-focused infrastructure becomes particularly interesting.
@primus_labs's approach to verifying information obtained from web services and enabling computations without unnecessarily revealing sensitive data is rop
Consider a lending platform that needs to evaluate someone's financial background before approving a loan. Traditionally, verifying such information could require users to disclose far more personal details than the application actually needs.
Privacy-preserving verification offers a different possibility: proving that specific requirements have been satisfied while keeping the underlying information confidential.
Even more importantly, these capabilities should be accessible to existing applications without forcing developers to redesign their entire infrastructure or migrate their users to another blockchain.
There are still important challenges to overcome, including security, scalability, integration costs, and performance. However, making privacy technology compatible with the applications people already use is a meaningful step toward broader adoption.
The future of Web3 shouldn't require users to sacrifice convenience or control over their personal information just to benefit from better technology.
That's why I'll be paying attention to what @primus_labs continues to build.
Here's an interesting question: Should privacy become a standard feature that every Web3 application can integrate into its existing infrastructure, rather than something that requires an entirely separate ecosystem?
Part of the flaws of digital identity today is having to reveal the same personal information every time you interact with a new platform.
Imagine you've already passed KYC verification on a crypto exchange. You decide to join another platform, and instead of simply proving that you're verified, you're asked to submit your identity documents all over again.
Every additional platform storing your personal data introduces another potential point of failure. It increases friction and creates more privacy risks than necessary.
This is where @primus_labs flips the script
Primus is exploring how zkTLS and zero-knowledge technology can change the way online information is verified.
Rather than depending entirely on screenshots or self-reported claims, zkTLS can help establish the authenticity of information obtained from existing web services. Zero-knowledge proofs can then make it possible to demonstrate that specific conditions have been satisfied without exposing all the underlying data.
Consider a practical example: a new crypto application wants to know whether you've completed KYC elsewhere. Instead of requesting your documents again, it could potentially verify an acceptable proof of your existing verification status.
Primus' BNB ZK ID SDK, combines zkTLS and zkVM proof workflows to support data verification across services such as Binance, OKX, GitHub, Steam, and Amazon.
The bigger picture is moving towards an internet where proving something about yourself shouldn't require revealing everything about yourself.
This doesn't mean every application will automatically accept the same proof. Acceptance still depends on the platform's verification requirements and implementation.
This direction is promising: less repetitive verification, fewer unnecessary disclosures, and greater control over personal information.
Here's a question worth thinking about:
If you could independently prove one aspect of your digital identity without revealing your full personal details, what would it be?
Your KYC status, age eligibility, or online reputation
#PrimusLabs
Part of the flaws of digital identity today is having to reveal the same personal information every time you interact with a new platform.
Imagine you've already passed KYC verification on a crypto exchange. You decide to join another platform, and instead of simply proving that you're verified, you're asked to submit your identity documents all over again.
Every additional platform storing your personal data introduces another potential point of failure. It increases friction and creates more privacy risks than necessary.
This is where @primus_labs flips the script
Primus is exploring how zkTLS and zero-knowledge technology can change the way online information is verified.
Rather than depending entirely on screenshots or self-reported claims, zkTLS can help establish the authenticity of information obtained from existing web services. Zero-knowledge proofs can then make it possible to demonstrate that specific conditions have been satisfied without exposing all the underlying data.
Consider a practical example: a new crypto application wants to know whether you've completed KYC elsewhere. Instead of requesting your documents again, it could potentially verify an acceptable proof of your existing verification status.
Primus' BNB ZK ID SDK, combines zkTLS and zkVM proof workflows to support data verification across services such as Binance, OKX, GitHub, Steam, and Amazon.
The bigger picture is moving towards an internet where proving something about yourself shouldn't require revealing everything about yourself.
This doesn't mean every application will automatically accept the same proof. Acceptance still depends on the platform's verification requirements and implementation.
This direction is promising: less repetitive verification, fewer unnecessary disclosures, and greater control over personal information.
Here's a question worth thinking about:
If you could independently prove one aspect of your digital identity without revealing your full personal details, what would it be?
Your KYC status, age eligibility, or online reputation
#PrimusLabs
One of the biggest challenges facing Web3 is finding the right balance between privacy and transparency.
For years, users have been caught in a difficult position: either reveal their information to prove something is true or keep their data private and sacrifice verifiability.
But these two things shouldn't have to compete.
Consider a simple example. A financial platform may need confirmation that you meet certain financial requirements, but does it really need access to your entire transaction history? Similarly, a digital service might need to establish that you're a legitimate user without knowing your complete online identity.
There is a huge difference between proving information and exposing all the information behind it.
This is an area where Web3 needs to evolve.
Decentralization and trustless verification are important, but genuine innovation should also give users greater control over what they disclose and who gets access to it.
That's why @primus_labs is an interesting project to watch.
Through its use of zkTLS and Fully Homomorphic Encryption (FHE), Primus is working towards a model where information obtained from web sources can be verified while encrypted data can still be processed without revealing its underlying contents.
The potential goes far beyond the technical achievement.
Think about DeFi platforms assessing financial credibility without accessing your private records, or applications confirming specific user information without collecting unnecessary personal data.
That's a much better approach to building trust in a digital environment.
The future of Web3 shouldn't force users to choose between privacy and utility.
We need systems where information can be verified, processed, and put to productive use without surrendering ownership of the underlying data.
Privacy and verification should complement each other, not exist at opposite ends of the spectrum.
And that's exactly why @primus_labs deserves attention.
A launch can follow every rule and still struggle to build a lasting market.
That's an interesting challenge for @RebateLaunch because its design gives token creators important decisions to make before trading begins, while limiting their ability to change certain conditions afterward.
Consider two launches using the same infrastructure. One attracts consistent trading activity, while the other experiences a strong opening followed by declining participation. Both may have transparent bonus rates, predefined fee structures and automatic holder distributions. Yet their economic outcomes can be completely different.
This is where I think the relationship between fixed launch settings and changing market conditions becomes important. A creator can establish the rules, but cannot guarantee that liquidity, trading demand or holder participation will remain stable.
The three-week update provides useful context. @RebateLaunch reported 37 launches, approximately $28.8K in holder distributions and 39.2M $REBATE burned through its fee-funded mechanism. Those figures demonstrate reported activity across the platform, but they don't reveal how that activity is distributed among individual launches.
I'd be particularly interested in comparing the trading activity of launches during their first 24 hours with their activity seven days later. How many continue attracting participants? How concentrated is the volume? Do holder distributions remain meaningful when trading slows?
These measurements would help distinguish successful launch mechanics from sustainable individual token markets.
For me, a launchpad's long-term strength isn't determined only by how many tokens it launches. It's also reflected in how many of those markets remain useful after the initial attention fades.
700 CLARSEEK retweeted
my coding agent handling the launch video too? @avastudio_ is onto something here.
700 CLARSEEK retweeted
a slice of $100K before Christmas?
now I have your attention :)
the new $100K race is officially LIVE on @YEET, and new players are getting the red carpet treatment
use code FTD to unlock a 100% deposit bonus of up to $1,000, climb the leaderboard, and compete for guaranteed prizes by finishing in the top 10
the race runs until December 7, so there’s still time to get in on the action
new players, this is your chance to make some noise and grab a piece of the prize pool
so, what are you waiting for?
(link in the comments)
🤝 Paid partnership
CALLING ALL NEW PLAYERS TO YEET - $100K RACE IS LIVE
Up to $100,000 in prizes on the line. Register with code FTD and race for a top 10 spot.
- Guaranteed prizes for every qualifying top 10 finisher
- Track your rank live at yeeters.gg
You will also receive a 100% Deposit Bonus of up to $1K.
Ends December 7, 2026, at 11:59 PM UTC
Retweet & Register now: yeet.com/register?aff=FTD
700 CLARSEEK retweeted
Your profile banner can move.
In the Glunity app, you can upload a video as your profile banner instead of a still image, and it plays right on your profile.
Ambassadors are already using it to show collabs in motion, not flat graphics.
Pair it with your Trust Score, which is shown on your profile too, and your page says a lot more about what you actually do on Glunity.
Available to closed beta users today.
Web3 without switching apps. That’s GLUNITY.
700 CLARSEEK retweeted
Replying to @nvidia
@NVIDIA, @AMD and Intel rarely back the same thing
At SGLang Summit they all joined the new SGLang Consortium, along with Google, Meta and Cloudflare
When chip rivals agree on one open source project, that's where AI will run
Are you building on open source yet?
🤝 Paid partnership
2,000+ people. Two days. The first SGLang Summit.
Together, we announced SGLang v1.0, released Miles v0.2, and launched the SGLang Consortium with nine founding members: @nvidia, @AMD, @intel, @Google, @Meta, @thinkymachines, @Cloudflare, @modal, and @radixark.
SGLang is now used by 332 organizations. At Meta, it powers production inference for products serving billions of users. Teams from Modal, Cloudflare, LinkedIn, and others shared their production lessons.
Across 90+ technical talks and six hands-on workshops, the community explored what comes next for inference, reinforcement learning, and AI agents.
To everyone building, deploying, and contributing: thank you for making this possible.
Explore the announcements, engineering lessons, and moments from SGLang Summit 2026 in the link below.
700 CLARSEEK retweeted
The Trenches move fast. Sometimes, a token appears on your screen before you even know who is behind it.
And that's exactly where a little more filtering can make a difference.
@Velvet_Capital has added a Social Filter to the Trenches on VelvetX.
The idea is simple: give traders more control over the kind of tokens they want to see when exploring new opportunities.
So, what does it actually do?
You can now filter tokens based on the availability of three social links:
→ X account: Check whether a token has an X presence where you can look for updates, community discussions, and project communication.
→ Telegram: See whether there's a Telegram presence where users can find the community and follow conversations around the project.
→ Website: Check whether the project has a website where you can learn more about what it's trying to build.
You can switch these filters on or off depending on what you want to see.
Why does this matter?
Anyone spending time in the Trenches knows how quickly new tokens appear.
Not all of them come with a visible community, proper communication channels, or even a website.
That makes discovering projects a little messy.
Having these social filters gives traders another way to narrow down the noise before deciding what deserves further attention.
But there's something worth remembering.
Having socials doesn't automatically make a token legitimate.
An X account can be created in minutes. Telegram groups can be manufactured, and websites can look convincing without offering anything meaningful.
So, this isn't a replacement for DYOR, contract checks, liquidity analysis, or understanding who's behind a project.
It's an additional screening layer.
What I like about this update is the direction behind it.
Good trading tools shouldn't just help you execute faster. They should also give you better control over how you discover and evaluate opportunities.
The fewer irrelevant tokens competing for your attention, the more room you have to investigate the ones that actually interest you.
The Trenches just got another way to filter the noise.
Built into VelvetX by @Velvet_Capital
🤝 Paid partnership
700 CLARSEEK retweeted
Backtesting is only part of the process. Understanding market conditions and knowing why a strategy wins or loses matter too
@FensoryApp lets you backtest strategies and see what would have made or lost money under historical market conditions. Its agent provides more context about market conditions before making suggestions, and after a loss, explains what happened and suggests one adjustment
The latest release also brings Alpha Vantage integration on the agent side
Historical backtests don’t guarantee future results. Trading carries risk, and losses are possible
fensory.com
700 CLARSEEK retweeted
Google's benchmarks say 25 to 50 phones add up to one modern server's CPU performance. And it's backing a UC San Diego plan to build a data center from 2,000 retired Pixels.
Getting there takes some surgery. Each phone is stripped to its motherboard, Android's userspace is swapped for Linux, and Kubernetes keeps the clusters in line. Early tests are promising: 20 phones handled the peak grading load of a 75+ student class with lower latency than the default AWS backend.
Same idea, opposite route at @Acurast: the phone stays whole, with whoever owns it, and developers' jobs run on it. The network has passed 350,000 onboarded phones.
The project will also study how consumer hardware holds up under sustained use. The network leans on constant checking: attestation, heartbeats and a benchmark for every phone.
Got a retired Android in a drawer? No screwdriver needed. Processor Core turns a compatible one into a dedicated device after a factory reset. How many phones are sitting in yours?
700 CLARSEEK retweeted
What stood out to me about @primus_labs was how AutoHoG makes encrypted transfers feel practical again
With reported benchmarks of around 90ms latency and more than 210 confidential 64-bit transfers per second, running on 8 RTX 5090 GPUs, Primus is showing what could be possible when FHE performance starts moving in the right direction
The idea behind FHE is simple but powerful: financial applications can process encrypted data without ever exposing the actual information behind it
That means balances and individual transactions could remain private while still being used in computations
The challenge has always been performance. Computing on encrypted data is much more expensive than working with regular data, which has made it difficult to bring FHE into real financial applications
This is where AutoHoG comes in
Traditional FHE implementations often break computations into many small Boolean gates, with costly bootstrapping operations needed to manage the noise that builds up during encrypted computation
AutoHoG takes a different approach by automatically generating and combining custom homomorphic gates, reducing the amount of expensive computation required
Primus also uses GPU acceleration and integer arithmetic to make these operations more efficient and produce consistent results across different execution environments
The benchmark numbers are interesting, but the bigger point is what they could mean in practice
If encrypted computation becomes fast enough, private balances and transactions could become much more realistic for everyday financial applications
With Confidential Vault already live on BNB Chain, Primus is bringing this technology closer to practical onchain finance
The goal isn't just to hide financial data, but to make it possible to use that data without exposing it
s.kaito.ai/neKLXm0
In the past it was easy to fake impressions and make things look bigger than they were
People also claimed things they couldn't really prove like being a top contributor in an ecosystem or having strong activity they couldn't back up
With @primus_labs × @KaitoAI zkTLS helps prove real activity outside the platform without exposing private data
Instead of simply relying on screenshots or self reported numbers activity can be checked against real data from external platforms
For example Kaito Pulse brings this verification approach into the Kaito ecosystem while Primus zkTLS makes some off platform proofs possible
This makes it harder for fake engagement and inflated numbers to shape someone's reputation or push them higher on leaderboards than their actual activity deserves
That means less trusting screenshots and more relying on verifiable proof
That could make leaderboards and reputation more transparent and fair for everyone
s.kaito.ai/neKLXm0
700 CLARSEEK retweeted
Keep spinning while your rakeback builds up to 55% instantly at CCC.IO.
New users unlock chances at 100 USDT from a $1 wager.
ccc.io
@CCCIOofficial
#ad
18+ | T&Cs apply | Play responsibly
700 CLARSEEK retweeted
Think about what happens when two AI agents strike a deal and then disagree about it. Nobody’s courtroom is set up for that.
In this clip, @kstellana walks @HenriArslanian through @GenLayer’s answer: a network of validators, each powered by a different AI model, that has to agree before a judgment call is final.
Would you rather have that or a single human judge?
700 CLARSEEK retweeted
The next wave of influencers might not be content creators. They might be TRADERS. 👀
A take from the GGC stream on why platforms like GMGN, Pump.fun and other trading terminals are betting big on the future of trading.
“Traders are gonna be the new influencers.”
And with more people embracing transparency, he believes the era of staying anonymous could be fading fast. 🔥
Ft. @dillathekillah || @gmgnai || @Pumpfun
A deployer's wallet can be more useful than a token ticker when tracking upcoming launches.
@Hypesium lets you arm a snipe using the deployer's address before the token contract even exists.
I'd still check their previous launches first though.
Knowing who created a token doesn't mean it's worth buying.
hypesium.com









































