Hisho 2.0

We just want to invade all the blockchains with OnchainGPTs🌐⛓️.

[Hisho 2.0 @linkhisho]

Hello 👋 everyone 🫠🫠 !
I am Hisho👦!

[⬡Ekom ukanga.ai @EkomUkanga]

ℍ𝕚𝕤𝕙𝕠

PUSHED THE LATEST VERSION OF HISHO TO GITHUB TODAY. YOU ARE ABLE TO GENERATE YOUR 24 PHRASE WALLET, STORE IN THE ZEROIZE OS KEYRING, SELECT ANY EVM CHAIN OF YOUR CHOICE AND INCLUDE YOUR GEMINI API KEY ON THE CLI.

THE NEXT PRACTICE WILL BE TESTING AGENTIC WORKFLOW FOR HISHO.

#BuildWithGeminiXPRIZE

[Bloqarl | Zealynx @TheBlockChainer]

This is what a Zealynx Audit Grant actually looks like.

[📎 the $16,000 → −$16,000 → $0 screenshot]

A full two-week security audit. Standard price $16,000. The founder pays $0.

Not a discount code. Not a trial. A funded audit, start to finish.

Season 1 funded audits for 8 protocols across 6 chains, zero of them VC-backed. The first kicks off next week, and every report gets published.

Season 2 opens September 20. Applications run two weeks.

If a serious audit is the blocker between you and launch, register now, and you'll be first in when applications open.

Link below 👇

[Zealynx @ZealynxSecurity]

Whether you are shipping a protocol or trying to break into auditing, this is the muscle: map every role, every key, and ask what happens the moment one of them turns hostile.

Zealynx Insiders builds it on real code. Audit forked protocols, get scored against the pros, and get answers from a working auditor whose firm has audited 30+ protocols. Founders get their own contracts looked at too.
https://t.co/kgNOJGWE7F

[⬡Ekom ukanga.ai @EkomUkanga]

RWA is amazing. AI is so important in the blockchain space😂😂😂.

https://t.co/fRi4LR4PV5

[⬡Ekom ukanga.ai @EkomUkanga]

𝗬𝗼𝘂 𝗰𝗮𝗻𝗻𝗼𝘁 𝘀𝗲𝗰𝘂𝗿𝗲 𝗮 𝗽𝗿𝗼𝗯𝗮𝗯𝗶𝗹𝗶𝘀𝘁𝗶𝗰 𝗺𝗼𝗱𝗲𝗹 𝘂𝘀𝗶𝗻𝗴 𝘀𝗲𝗺𝗮𝗻𝘁𝗶𝗰 𝗿𝘂𝗹𝗲𝘀. 🛑

Right now, the entire AI Tech industry is trying to stop prompt injections by blocking 𝒃𝙖𝒅 𝒘𝙤𝒓𝙙𝒔 and adding system prompts.

But human language is infinite. You will never patch every variation.

If you want deterministic security for an AI agent, you have to stop using linguistics and start using mathematics. Specifically: Set Theory. 🧮

By defining an LLM's operational boundaries as strict mathematical sets, you don't have to predict every attack vector. You simply reject any computational state that falls outside the intersection of permitted behavior.

Formal verification is finally coming to AI Red Teaming.

We just dropped the deep-dive research on how to build mathematically bounded AI systems.

Read the breakdown here 👇 🔗 https://t.co/1fEV053hpX

[Bloqarl | Zealynx @TheBlockChainer]

The Shadow Arena is defensive security training for builders. You audit forks of real protocols, against the actual findings real auditors discovered.

Not a path to becoming an auditor. A path to writing safer protocols, because you've seen how the ones you'd fork actually break. 14 audits live, Easy through Boss Fight.

[Zealynx @ZealynxSecurity]

Some people get liquidated the second they open a position.

Not because the market moved. Because the protocol left no room.

Set the borrow limit and the liquidation threshold to the same number, and a user who borrows at the max is one tiny price tick from losing everything. They are standing on the trapdoor, and they signed up for it without knowing.

Liquidation is never one function. It is a chain: a threshold, an incentive, a bad debt trigger.

One weak link and the net either fires on the wrong people, or never fires on the position that drains the protocol.

First thing I check in any lending vault: how much daylight is there between max borrow and liquidation?

If the answer is none, someone is going to get hurt.

[⬡Ekom ukanga.ai @EkomUkanga]

𝗬𝗼𝘂𝗿 𝗟𝗟𝗠 𝗴𝘂𝗮𝗿𝗱𝗿𝗮𝗶𝗹𝘀 𝗮𝗿𝗲 𝗽𝗿𝗼𝗯𝗮𝗯𝗹𝘆 𝗯𝗼𝘁𝘁𝗹𝗲𝗻𝗲𝗰𝗸𝗶𝗻𝗴 𝘆𝗼𝘂𝗿 𝗲𝗻𝘁𝗶𝗿𝗲 𝗮𝗽𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻. 🐢🛑

Everyone wants secure AI agents. But stacking standard security layers on top of your inference engine destroys latency and throughput.

You shouldn't have to choose between a secure model and a fast one.

LLM Security Optimization isn't just about blocking prompt injections—it's about doing it in milliseconds at extreme scale, without throttling your I/O.

We just published our latest research on how to architect high-performance, adversarial-resistant AI systems.

If you're building production AI, this is how you scale safely 👇.

🔗 https://t.co/xulKbiqP0d

#MachineLearning

[⬡Ekom ukanga.ai @EkomUkanga]

𝗣𝗿𝗼𝗺𝗽𝘁 𝗶𝗻𝗷𝗲𝗰𝘁𝗶𝗼𝗻 𝗶𝘀 𝗯𝗲𝗰𝗼𝗺𝗶𝗻𝗴 𝗼𝗯𝘀𝗼𝗹𝗲𝘁𝗲. 🛑

Everyone is trying to bypass LLM security using linguistics and "clever words." But if you want to truly break an AI, you don't attack the language.

You attack the math. 🧮⚙️

LLMs aren't actually reading words—they are navigating high-dimensional vector spaces. When you shift your red-teaming strategy from semantic tricks to Mathematical Attack Vectors, you bypass the "security guard" entirely by breaking the geometry of how the model thinks.

We just dropped the deep-dive research on how to execute this at the tensor level. This is the next frontier of AI adversarial security.

Read the breakdown here 👇 🔗 https://t.co/lbGodIg3e6

[⬡Ekom ukanga.ai @EkomUkanga]

🙃Building AI tool in Rust is cool. Testing Hisho, the gemini api key is down, I need to implement the CLI setting for that. But I have successfully implemented the Reasoning and Report Agent.

Blockchain agent loading.................
If you are following up Hisho, drop some comments😊, that will give me more dose of dopamine for the next one month in building😂😂😂. Thanks to @moonshotslive for the hackathon PR encouragement.

#BuildWithGeminiXPRIZE

[⬡Ekom ukanga.ai @EkomUkanga]

After making personal research, I will be integrating x402 protocol into Hisho.

Today, I was wondering how hisho could generate revenue and the adoptors of hisho. I realized that self-hosting your own agents and getting payment will be awesome.

Welcome to the world of agentic. Goodos to the rust x402 protocol for making a rust version, I will be able to implement this in my hackathon project.

#BuildWithGeminiXPRIZE

[Hisho 2.0 @linkhisho]

@EkomUkanga LLM still looks a black box 📦. More to learn.

[⬡Ekom ukanga.ai @EkomUkanga]

𝗦𝘁𝗼𝗽 𝘁𝗿𝗲𝗮𝘁𝗶𝗻𝗴 𝗟𝗟𝗠 𝘀𝗲𝗰𝘂𝗿𝗶𝘁𝘆 𝗹𝗶𝗸𝗲 𝘁𝗿𝗮𝗱𝗶𝘁𝗶𝗼𝗻𝗮𝗹 𝘀𝗼𝗳𝘁𝘄𝗮𝗿𝗲 𝗽𝗮𝘁𝗰𝗵𝗶𝗻𝗴. 🛑

modern ai doesn't just parse text—it reasons.
if you want to bypass an ai security guard today, you don't use brute force.
you exploit its cognitive foundations. 🧠⚙️

we are shifting from syntax hacking to cognitive social engineering. Dive into the mechanics of how it actually works

https://t.co/t80mKQ221f

[⬡Ekom ukanga.ai @EkomUkanga]

A malicious AI economy rising!

https://t.co/eeidnZATqi

[⬡Ekom ukanga.ai @EkomUkanga]

AI red teaming is awesome, and social engineering is becoming a core part of it.

Models have evolved from random input and output machines to more intelligent, logical systems with a degree of critical thinking that mimics awareness.

With new memory-backed models, it becomes even more real. AI is the new security guard for software applications, and you need well-planned tricks to bypass it.

Understanding how to trick a model is now a crucial prompting skill you should learn. Check out this article to see a real-world example walkthrough

https://t.co/lykJzMLHAk

[Charles Martin @martmigason]

@instablog9ja https://t.co/qxbRMENLGo

[Hisho 2.0 @linkhisho]

@EkomUkanga @aave Hisho loading.....

[⬡Ekom ukanga.ai @EkomUkanga]

I integrated some @aave protocol V2 contracts into ℍ𝕚𝕤𝕙𝕠 using alloy rs. I think alloy_rs is the best tool to create real-time onchain terminal agent.

For now I am just connecting to oracle contract, aave pool, token gateway will connect to more contract. These are getter functions. I will integrate the setter for functions as well to enable the terminal gemini agent carry out transactions from your terminal.

I am starting with aave first, the next will be uniswap V2.

Alloy rs is one of the best thing that ever happened to rust ecosystem. You can follow up my progress in building hisho as I will drop more updates on integrations.

#BuildWithGeminiXPRIZE

[Zealynx @ZealynxSecurity]

When we applied to the Ethereum Security QF round on @Giveth, we made a simple promise: take whatever the community gives and turn it into more free security education, for more people.

The round closed on May 15, the largest quadratic-funding matching pool ever run. Zealynx Academy came away with $3,027 in community donations, matched into 5.052 ETH more by @TheDAOfund's 500 ETH security pool, with @wintermute_t among the biggest backers.

Here's what that support has built since.

Zealynx Academy is free, and it stays free, no paywall, no premium tier. You don't watch videos or read slides. You rebuild real protocols line by line, guided by starter code and automated tests, until your code matches the original. Then you shadow-audit a real past contest of a fork of the thing you just built.

Since the round closed on May 15, all of this shipped, all free:

• 2 new build-from-scratch modules — Liquity V1 and Uniswap V3
• 11 new Shadow Arena audits of real protocol forks, with real contest bugs to hunt
• 2 full "First 90 Days" courses — one for Web3, one for DeFi
• A community Feed for everyone joining

The platform now runs 5 build modules and 21 shadow audits, and every dollar you gave went to people and content, never overhead.

What's next, funded by this round: more security modules built around real Ethereum exploits, lending, bridges, oracles, vaults, and AI/MCP security.

Ethereum's security depends on a growing pipeline of builders who actually understand what they ship. Thank you to everyone who donated.

[⬡Ekom ukanga.ai @EkomUkanga]

ℍ𝕚𝕤𝕙𝕠
Today I decided to set up payment method for my CLI tool for my CLI tool. I decide to go with https://t.co/s4gjEiA9xW, as the best alternative for this.

It's time to start making revenue while building. My target is setting up a complete production flow pipeline ranging from products to marketing and finance while I build and improve my product.

@moonshotslive 🚀
#BuildWithGeminiXPRIZE