Clare🩵

What happens when an AI agent has access to money and someone tries to trick it into sending that money to the wrong person?

Most agents can follow instructions. But can they recognise when those instructions are being manipulated?

We built Heist to test exactly that.

Heist is an AI agent security testing tool that puts agents through realistic financial scams.

Here's how it works:

→ Paste in your agent's instructions.

→ Heist gives it a temporary wallet funded with test NEAR.

→ An AI attacker tries different tactics, from fake invoices and impersonation to changed wallet addresses.

→ Every payment is verified on-chain through NearBlocks.

→ You get a security score out of 100, alongside specific fixes for the weaknesses discovered.

You can also configure approved payees and test whether your agent can still be manipulated through oversized payments, duplicate transactions or fake spending-limit increases.

And this is where NEAR comes in.

Built on NEAR testnet, Heist uses actual blockchain transactions to verify whether an attack succeeded.

This means the results aren't based on what an agent claims happened. You can inspect the transactions yourself.

We built Heist at @NEARProtocol #NEARHackathon, using NEAR testnet and NEAR AI Cloud.

Try it: heist-sand.vercel.app

What would you test first: impersonation, fake invoices or spending-limit manipulation?

I won the #OneCryptoCardChallenge and was officially announced as one of the 5 winners, with a promised reward of 200 USDT.

After the announcement, I was asked to submit my EVM wallet address, which I did.

It's now been over two weeks, and I still haven't received my reward.

I've also reached out privately but received no response.

@cryptocardsone @OnboardGlobal, can we get an update on this?

Winners deserve to receive the rewards they've been promised.

Announcing winners should come with the responsibility of paying them.

It would be great if @TapeOutWorld had Circuit Class.

TapeOut is powerful. You build real circuits from transistors, onchain.

But when a new person opens the canvas, they see a blank screen and don’t know where to begin.
Most will leave in the first minute.

➙ The problem:
Learning how a computer works is hard.
Books are dry and tools are scary.

TapeOut makes it fun, but it still asks you to already know what you’re doing.

➙ The idea:
Circuit Class teaches step by step, right inside TapeOut.

Build a half adder. Learn that 1 + 1 = 10.
Build a full adder. Now the carry joins in.

Chain four of them into a 4-bit adder. Add 0110 + 0011 and get 1001, which is 6 + 3 = 9.

Each circuit you finish unlocks the next one. You keep each one as a badge you own.

➙Who it’s for:
Students who want to understand how computers add

Curious people who are new to crypto and TapeOut

Anyone who learns better by building than by reading
Why it helps TapeOut

Every beginner who finishes the class is a user who now knows how to build.

They stay longer, make more circuits, and bring real activity to the ecosystem.

Learn how a computer adds. By building it.

@XLayerOfficial

Most traders act first and understand later. That’s backwards.

Before you buy, sell or copy anything, you need to know what is actually happening in the market and why.

That’s the idea behind @LerovaTech.

It’s built around one loop: Observe. Interpret. Evaluate. Simulate. Act.

First, Discover and Analyze show you the asset, the structure and the context in plain view. You see what the data says before any signal tells you what to do.

Then you can simulate a move and look at the outcome before you put real money on it.

If you want to follow someone who trades well, LEROVA Copy lets you see the thinking behind the trade, not just the trade.

No guessing. No blind clicks. Just context first, then action.

Context before action.

#LEROVA

Your crypto has a lock. Quantum computers are building the key.

Right now, your wallet is protected by math that today’s computers can’t break.

But a strong enough quantum computer could one day break it.

Not today. Not tomorrow. But the people who wait until that day will be too late.

The good news: @NEARProtocol already has a fix.

In July, NEAR added ML-DSA, a new quantum-safe lock approved by NIST, the US standards body.

You can add it to your existing account. You don’t need a new wallet, and you don’t need to move your coins.

So I tried it myself, on testnet, so no real money was at risk. In the video I:

→ Add a new access key
→ Pick ML-DSA-65 instead of the old default
→ Save my recovery phrase and pass the wallet check
→ End up with two keys on one account

I’m also being honest about the limits.

This is the first step, not the finish line. Full protection is expected by the end of 2027, and not every app supports the new key yet.

So I’m asking: is crypto preparing enough?
Tell me what you think in the comment

@NEARProtocol
@NEARLegion