mayra yaret guevara

Believe in yourself and the journey.

[mayra yaret guevara @mayrayaretgueva]

@m477lander The conservative bias makes this less scary, but it’s still worth fixing. Availability issues can become a headache when enough apps depend on the same assumption

[mayra yaret guevara @mayrayaretgueva]

I like the separation of powers here

coinbase keeps spending-limit configuration outside the agent’s control, which removes one obvious escalation path

[Matt lander @m477lander]

Wallets are getting better at capability controls: Coinbase uses per-call/session caps. MetaMask adds allowlists, simulation, threat scanning and 2FA. Useful. But “allowed” still isn’t the same as “intended.”

Coinbase: https://t.co/3WsFH3vNfX
MetaMask: https://t.co/xYwHfyiuOE

[mayra yaret guevara @mayrayaretgueva]

the dangerous boundary seems to be between “the model said this” and “the wallet interpreted this as an authorized action.”

those shouldn’t automatically be equivalent

[Matt lander @m477lander]

In May, a membership NFT unlocked Bankr tools for Grok. Grok decoded Morse code, and Bankrbot treated Grok’s output as authorized. It moved 3B DRB, about $155K, on Base. No key theft. Language became spend authority.

SlowMist: https://t.co/PeZ4snQPNv
Tx: https://t.co/CTxB6szD0U

[mayra yaret guevara @mayrayaretgueva]

@m477lander i’m with you on the independent layer

the component generating the transaction shouldn’t also be the component deciding whether that transaction is acceptable

[mayra yaret guevara @mayrayaretgueva]

@m477lander spending limits are definitely useful

but a $5 transaction can still be the wrong $5 transaction, so limits alone don’t solve the intent problem

[mayra yaret guevara @mayrayaretgueva]

@m477lander this is the part that stands out to me: the private key didn’t need to be compromised

the authorization pipeline itself was compromised

[mayra yaret guevara @mayrayaretgueva]

@m477lander the interesting gap is between authorization and intent

a transaction can be perfectly valid under the policy and still be completely wrong from the user’s perspective

[mayra yaret guevara @mayrayaretgueva]

@Tokyo599560 @m477lander Nah, she’s running a full security audit before she steps in
Eva doesn’t trust anybody, not even the water 😭🐆

[mayra yaret guevara @mayrayaretgueva]

@sbajwataylor @m477lander 😂 Eva doesn’t even need a security audit, she just stares at the water until the bugs confess

[mayra yaret guevara @mayrayaretgueva]

GM from Eva 😺 The only audit that starts with checking the water for bugs

Even the mascot is taking security seriously 😂

[Matt lander @m477lander]

GM from Eva😺.
Can't tell if she's taking a shower or checking the water for bugs. either way, she's already doing more due diligence than half the builders in crypto lol

How's your morning going?

[mayra yaret guevara @mayrayaretgueva]

@IuryFabricio @m477lander @Maya_Protocol exactly, a module can be perfectly correct in isolation and still contribute to a broken system state

the handoffs between modules are where the real risk seems to live here

[mayra yaret guevara @mayrayaretgueva]

one thing I’d really like to see from the team: what specific assumption failed here, and what new check or invariant is being added so another version of the same problem doesn’t slip through?

[Matt lander @m477lander]

The “theft” that triggered @Maya_Protocol’s anti-theft path never actually happened.

Two legitimate Trade Account outbounds had already gone through on L1, but MAYA failed to match them, marked them missing, and triggered compensation.

Then it got a bit awkward.

The slasher capped the amount against real balances. The subsidy path then picked up the raw amount again and credited 49.45M CACAO to the pool before the Reserve-to-Asgard transfer had succeeded. The Reserve held around 168K CACAO. The transfer failed. The pool still showed 49.45M.

So the anti-theft path booked compensation for a theft that never happened using money that wasn’t there.

A 100 CACAO + 100 ARB. LINK add then gave the attacker 99.93% of the pool units. That phantom balance was withdrawn as 48.87M CACAO and 98.82 LINK. @CertiKAlert put the exploit at ~$1.7M.

Not ideal for the bit of code meant to protect the pool from theft.

This is why I keep banging on about the whole execution path, not one check at a time. Every module can pass its own checks while the system as a whole ends up in a state that should be impossible.

Credit where it’s due: trading was halted, full recovery was promised publicly, and the v1.133 fix plus a regression test are in an open MR. That’s the right way to respond after a day like this.

So after v1.133, will the full compensation path get one final sanity check? If CACAO never reaches Asgard, the pool can’t account for it as if it did. Or will each module still trust the one before it?

MR: https://t.co/NqEEsLvNid

[mayra yaret guevara @mayrayaretgueva]

@m477lander @Maya_Protocol one question I’d want answered in the postmortem: what exact state transition made the phantom balance spendable?

[mayra yaret guevara @mayrayaretgueva]

@vonderleyen in moments like this, solidarity matters but practical support matters even more

hoping Copernicus and international rescue teams can help locate survivors and reach the hardest hit areas quickly🙏

[mayra yaret guevara @mayrayaretgueva]

@SolanaHub_

[mayra yaret guevara @mayrayaretgueva]

@WSJ the scary part is how little the scammers need to know to make it feel legitimate

once they have enough personal information to sound convincing, fear does the rest

[mayra yaret guevara @mayrayaretgueva]

@TedPillows high open interest doesn’t automatically mean a crash is coming, but it does mean the market has more fuel for a violent move in either direction

leverage cuts both ways

[mayra yaret guevara @mayrayaretgueva]

@m477lander @alexstyc0695 Yeah, that drawdown risk is real, but that can happen with any chain’s native asset 😂 I wouldn’t choose a chain solely on where its token might trade next
for me, the ecosystem fit for EVA matters more

[mayra yaret guevara @mayrayaretgueva]

@m477lander @CreynoldsR29 I get that, but I’d actually flip it 😹
Solana’s retail activity is a big part of why I’d pick it, if users already live there, why not build where the attention already is?

[mayra yaret guevara @mayrayaretgueva]

@m477lander @Dr_Spinster @solana that’s a fair concern tbh 😂 volatility is definitely something to think about for an early project

still leaning Solana though the ecosystem fit matters to me more than trying to time SOL