
@hackenproof ex Community Manager intern Smart Contract Auditor https://teapawt.net/t.co/C2smHSJZVN
Joined September 2020
- Tweets536
- Following926
- Followers210
- Likes2.7K
I first heard about the term TEE through @bluethroat_labs . Reading their handbook pushed me to explore TEE security in blockchain.
My first article looks at a published attestation failure.
dev.to/khaye/in-tee-attestat…
10
225
Elizabeth not the Queen retweeted
Excited to partner with @VittoStack and the Jailbroken community to support AI security research. 🛡️
We're providing $25 in free OrcaRouter API credits for researchers to red team LLMs, evaluate jailbreaks, and test AI systems using 200+ models through one API.
Break things responsibly. 🐋
Today we're launching two post-trained, uncensored open-weight models on OrcaRouter:
• Gemma 4-26B-A4B-Uncensored (Balanced)
• Qwen 3.6-35B-A3B-Uncensored (Aggressive)
These models are designed for legitimate security research, AI safety evaluation, red teaming, jailbreak robustness testing, and alignment research.
To prevent abuse, requests still pass through a fail-closed safety classifier for high-severity categories such as CSAM (Child Sexual Abuse Material) and CBRN (Chemical, Biological, Radiological, and Nuclear) and access to these models are gated.
Why?
We believe the future of AI security depends on giving defenders access to frontier open-source cyber capabilities while maintaining strong safeguards against the highest-risk abuse.
Choose your model:
🛡️ Balanced — keeps more of the original behavior while removing unnecessary refusals.
⚡ Aggressive — fully unlocked for maximum testing coverage.
Available now:
Gemma 4-26B-A4B-Uncensored (Balanced) → orcarouter.ai/models/obsidia…
Qwen 3.6-35B-A3B-Uncensored (Aggressive) → orcarouter.ai/models/obsidia…
A good way to start your smart contract auditing journey
Audit Arena Week 23 is live ⚔️
Every week, we see builders, auditors, and learners come together with one shared goal:
To get better.
Not by talking.
Not by pretending.
But by showing up, reading code, finding bugs, asking questions, and improving one submission at a time.
That’s what Radcipher is building with Audit Arena, a place where effort compounds and real security skills are forged.
And this is only the beginning.
More is coming.
More challenges. More learning. More opportunities. More value for the security community.
We are here to produce, to build, and to keep pushing the standard higher.
Week 23 is now open.
Step in. Compete. Learn. Sharpen your edge.
The Arena is live 🏆
2
147
Elizabeth not the Queen retweeted
There's also a heatmap for projects, where brighter areas highlight those most likely to contain bugs. You need to train your eye to spot them.
Replying to @Parth0x108
That's the hard part.
There's a reason why almost anyone can solve "find-the-bug" challenges, but very few people can find live vulnerabilities in the wild.
CTFs have bounded domains and usually one clear, proven solution. Live exploits, on the other hand, are unbounded -- we don’t know where they are, or even if they exist at all.
That said, there are hot zones in codebases -- areas where developers are more likely to make mistakes.
There are also danger zones -- places where even a small error would be fatal.
Once you’ve worked with many similar codebases, languages, patterns, and technologies, you start to see those hot spots instinctively.
Hunting is like reading a heatmap.
Elizabeth not the Queen retweeted
STOP USING AI FOR EVERYTHING. STOP ADDING AI FEATURES EVERYWHERE.
The levels of sloppification are becoming wild.
This is so tiring. Be human.
Elizabeth not the Queen retweeted
Replying to @lonelysloth_sec
One day people will realize that their intelligent agent doesn't reason at all, as it's a glorified static analyzer.
It sees some patterns a million times and says "Yeah, this is a bug for sure!"
Then it chains two patterns it saw a million times and says "I found this issue via deep thinking!"
It may work for the known issues it has crawled over and over, as long as they're in its feed.
Once researchers stop feeding it new stuff, it will become blind.
There's also another plausible scenario, where the AI is fed with all the slop it produces -- since people stop writing quality articles and papers, and it enters a cycle of slop reinforcement learning.
Elizabeth not the Queen retweeted
Audit Arena Week 19 is LIVE : Referral Echo Sale
A referral-based token sale contract just dropped.
LPs add/remove liquidity. Traders swap tokens. Somewhere in the logic - bugs are hiding.
Your mission: find them, write the report, submit before the week ends.
100+ researchers competing. Leaderboard is tight. Every point matters.
Think you can catch what others miss? 👇 🔗 radcipher.com/competition
Elizabeth not the Queen retweeted
Cognitive Demoralization.
It's a form of abuse. Or psychological warfare. And people are trying to do that to you on an industrial scale.
Convincing people their intelligence is useless, they should just give up and accept they are cognitively hopeless in the new world and they should just consume the thoughts spoon fed from machines owned by a few big corporations.
If you're young and struggling about the future and what's the meaning of working and studying, dont fall for this trap:
"What is the point of even checking what the machine produces! It's too complicated for us to understand! We should just trust whatever the machine gives us, and give it permission to do everything for us. Giving up thinking is our only way to survive."
Homo Sapiens have been on this Earth for hundreds of thousands of years.
Ours ancestors evaded predators, became predators, built tools, invented theories, machines, populated the entire world, all sorts of climates and endless challenges, generation after generation.
We invented the machines. We produced -- without being trained to -- all the data the machines train on. Nobody gave us millions of examples of math proofs before we invented the very idea of what a math proof is.
We weren't trained or evolved to go to space -- or build cities, or cure diseases, or write poetry -- and we did it anyway.
The machines are just faint echoes of our own intelligence.
****There's nothing a fucking machine can do that is beyond your cognitive ability.****
Don't let anyone convince you otherwise.
You have the same biological machinery as Einstein or Von Neumann. You descend from thousands of generations of the people who won the incredibly hard game of survival in a threatening, immensely complex, and ever-changing environment.
Have faith in your intellectual ability to make sense of the world no matter what it throws at you.
Anyone who says "wow! how did the machine do this? it's super human! nobody understand how they do it!" is either proudly announcing their own stupidity or deliberately gaslighting you to question your own value as a human being.
The idea is just a stupid piece of propaganda intended to turn you into a passive consumer -- or to take you out of the competition.
The tools are yours to use -- or not use -- based on your own evaluation of what is good for yourself.
"The future is AI doing everything! If you keep doing things you'll lose."
"The future is humans+AI doing everything! If you try to do anything without AI you'll lose!"
Just two variations of the same lie.
***Nobody knows shit about the future.***
Your future is yours to create.
All you need is to not give up thinking for yourself.
Do what you love. Get good at it.
And when anyone tries to convince you it's useless -- just take note. That person isn't your friend.
Yeay, I was awarded for a valid submission on @HackenProof hackenproof.com/ #hackenproofed #bugbounty
Elizabeth not the Queen retweeted
I read every audit report in my 2026 dataset. Every upgrade. Every ignored finding.
The contracts were clean on paper. Security got worse.
97 hacks last year. 68 already this year as of May 5, with Ekubo latest (in my dataset). Only lower because we haven’t had a 1B hit yet.
I’ve been looking at hacked protocols.
Who was audited. Who wasn’t.
I also tracked TVL loss by group, the attack vectors and impact, and which auditors signed off on the worst code. Was it @cyfrin? @peckshield? @CertiK? @MixBytes? @pashov? ToB? who exactly??
My graph is ready
Elizabeth not the Queen retweeted
3 months ago, we launched Audit Arena.
A place where anyone interested in Web3 security could show up, join the challenge, and sharpen their skills.
What started as a simple idea has grown into something we are genuinely proud of.
From competing in @immunefi @sherlockdefi @CodeHawks @cantinasecurity @HackenProof …
to building our own arena for researchers to practice, improve, and prove themselves.
It has been a beautiful journey so far.
Here is what the first 3 months look like:
- 100+ researchers joined
- 1000+ submissions made
- 100+ bugs found
But beyond the numbers, the most meaningful part has been watching people grow.
Researchers getting better week after week.
People showing up consistently.
Skills becoming sharper.
Confidence getting stronger.
That is what this was really built for.
Audit Arena is not just about competition.
It is about creating a place where talent can grow, where people can learn by doing, and where the next strong security researchers can be shaped through real practice.
We are still early.
But we are excited for what this can become.
And we are just getting started.
Elizabeth not the Queen retweeted
Applications for DevRel Uni Cohort 7 are OPEN 🚨
Deadline: April 25th.
If you're building in AI, Web3, or DevRel, this one's for you 🧵
Dive into a hands-on program featuring sessions led by @dabit8, @nutlope, @francescoswiss, @buzea200, @PSkinnerTech.
Elizabeth not the Queen retweeted
The @battlechain testnet is now LIVE.
Come enter the ultimate red-team platform.
Give us feedback so we can launch mainnet very soon, and fix web3 security.
Elizabeth not the Queen retweeted
The Flow Heist
Flow Network Exploit Post-mortem
On December 27, 2025, an attacker exploited a vulnerability in the Flow network to counterfeit tokens, extracting approximately $3.9 million USD across bridges. No existing user balances were accessed or compromised. The attack duplicated assets but did not touch legitimate holdings, with the vast majority of counterfeit assets being contained onchain or frozen by exchange partners before they could be liquidated. Network validators have ratified a decentralized governance action authorizing the permanent destruction of 100% of counterfeit assets. The network resumed operations on December 29th and is operating as expected with full transaction history preserved.
Attack Vector
The attack demonstrated significant technical sophistication. The attacker deployed over 40 malicious smart contracts in a coordinated sequence, exploiting a three-part attack chain:
1) Attachment import validation bypass
2) Circumvention of defensive checks on built-in types
3) Exploitation of contract initializer semantics.
The root cause was a type confusion vulnerability in the Cadence runtime (v1.8.8), now patched (v1.8.9 and later). The flaw allowed the attacker to disguise a protected asset (which should be non-copyable) as a standard data structure (which can be copied), bypassing the runtime's safety checks and enabling token counterfeiting.
Remediation
Beyond bridging assets out of Flow, the attacker attempted to deposit counterfeit FLOW across several CEX, with many exchange partners freezing the deposits upon receipt due to the abnormal size and internal AML protocols. Approximately 50% of those counterfeit FLOW deposits have already been returned by cooperative exchange partners (OKX, Gate, MEXC) and destroyed, and the Foundation is actively coordinating with remaining exchanges.
The network was restored on December 29, 2025, via an Isolated Recovery Plan that preserves all legitimate transaction history. This approach was selected following ecosystem-wide consultation with exchanges, bridge operators, and infrastructure partners. The Isolated Recovery Plan was chosen specifically to avoid reconciliation risk for off-chain custodial systems and cross-chain protocols maintaining independent state.
The Foundation is cooperating with blockchain forensic partners including zeroShadow and Find Labs and relevant law-enforcement authorities to support ongoing investigations.
Full technical details of the vulnerability, exploit mechanism, forensic analysis, and remediation architecture are linked in the comment.
Please watch this. Share it. Yell about it.
FORCE YOUR VC’s, FRIENDS, AND “DEFI” FOUNDERS TO WATCH IT.
They are going nuts rn on the OG BTC guys.
AND they literally just landed another $30m++ on Solana.
I am so fucking tired of these motherfuckers owning this industry. 😭
NEW PODCAST W/ @tayvano_ How North Korean Hackers Stole $300M+ Via Telegram w/ Taylor Monahan
Taylor joins @AsILayHodling to discuss the Lazarus Group’s Telegram scam SWEEPING crypto groupchats.
⚡️ Over +$300 million stolen
⚡️On-device wallets most at risk
⚡️Funds traced to NK's infamous Lazarus Group






























