Niknam

خوندن این پست برای هرکی تو کریپتو هستش از نون شب واجب‌تره

Your newest AI security engineer has just arrived. Cygent isn't just an AI bug scanner; it learns your codebase, works alongside your team, and writes the PRs to fix them. - Finds bugs - Gets on calls - Schedules tasks - Proactive security advice All in your chat interface 🧵

.@cyfrin has been working tirelessly to shift security left, into developers' hands, since day 1. - Building tools like @SoloditOfficial, Aderyn, Moccasin - Teaching people how to become auditors at @CyfrinUpdraft - Doing conventional audits And we just took another step

1/ We're a quarter into 2026, and crypto hiring looks nothing like prior cycles. We just published our latest Talent Insights Report breaking down how we got here and what it means for founders and talent teams. 👇

90% of blockchain dev interviews are just these 7 concepts repeated:

I think a lot of devs prepare for audits the wrong way. They look for more checklists. More tool outputs. More “common findings”. What usually matters earlier is whether you can actually reason through your own protocol: - what users do, - what the system assumes, - where flows break, - what an attacker would notice first. That is the kind of thinking I want to make more accessible with Shadow Arena. It is a place to get reps on security reasoning instead of just consuming more theory. academy.zealynx.io/shadow-arena

🚨Here is the perfect opportunity if you have 1+ year of web3 security experience. > Junior Web3 Auditor - Certik > $102-180k annual compensation > Pretty easy requirements jobs.lever.co/certik/478ab0c…

خوندن این پست برای هرکی تو کریپتو هستش از نون شب واجب‌تره

i was going to write this today. Happy to see that someone else did it.

If I was starting auditing right now here are the things I would do:

"Everything we just said is probably going to be outdated 5 minutes after this panel." @The3D_ (@aave) · @PatrickAlphaC (@cyfrin) · @PestoPoppa (@gnosis_) Mod: Peter Kecman (@Nethermind) By 2027, will we be auditing code, or auditing the AI agents that wrote it?

@hyperbridge, you said I should hack you, yh? x.com/hyperbridge/st… Well, unfortunately, I did not hack you, but I have a list of bugs, and this was the 3rd on my list. @seunlanlege It is better that you pause this bridge indefinitely.

> Spend $5-$10/month > Get a fully functional analyst > Keep tab on positions/holdings > Do deep research > Flag opportunities > Identify risks > It remembers & learns with you everyday ​ Result = saves more than an hour every day / replaces multiple workflows with 1

I’m not even surprised. I kept tagging @Polkadot and telling them that @hyperbridge was a scam. They kept ignoring. The so called Web3 philosopher has shown them his real color. @seunlanlege my head don catch you. I know you’re behind the hack . Don’t masquerade behind hackers. You wan tear another GLE to go blockfest for South Africa. Everyone who gave that guy a platform needs to be heavily scrutinized. Even @channelstv took him to the public eye and endorsers him and now he has demonstrated what he came for.

🚨 New day, new hack: @hyperbridge - $237K loss. An attacker exploited a vulnerability in Hyperbridge’s Ethereum gateway contract to mint 1B bridged Polkadot tokens and dump them for ~108 ETH. The team has paused all bridging.

AI is completely rewiring smart contract security in 2026. But pasting code into Claude won't print you money. Here is the actionable roadmap to build a dominant position in AI-assisted auditing before the window closes 🧵👇

So glad to have people using Zealynx Academy to improve their Smart Contract Security awareness. This is perfect for any Solidity devs that seek some advanced knowledge to improve the quality of their code. academy.zealynx.io

People think learning AI takes months. It's really just a couple of hours. And I wrote 17 free guides to start right away: Claude 101: ruben.substack.com/p/claude Claude Code: ruben.substack.com/p/claude-code Claude Skills: ruben.substack.com/p/claude-skills Nano banana 2: ruben.substack.com/p/banana-2-3bd Claude in Excel: ruben.substack.com/p/ai-couldnt-d… Best AI for Search: ruben.substack.com/p/grok-420 1M followers with AI: ruben.substack.com/p/1000000 Claude for your team: ruben.substack.com/p/claude-for-t… No prompt saves you: ruben.substack.com/p/magic AI Slides (PPT in 2026): ruben.substack.com/p/powerpoint Set up Claude Cowork: ruben.substack.com/p/claude-cowor… Claude to sound like you: ruben.substack.com/p/i-am-just-a-… Claude interactive charts: ruben.substack.com/p/claude-charts Claude as your computer: ruben.substack.com/p/claude-compu… Claude Cowork + Project: ruben.substack.com/p/claude-cowor… You're an AI workaholic: ruben.substack.com/p/ai-holic Setup AI before prompting: ruben.substack.com/p/how-to-bette… ___ 1. Save this list for later (three dots, top right). 2. Share it with a friend by ♻️ reposting this image. 3. Subscribe to my free newsletter: how-to-ai.guide.

In 8 weeks we went from 5% to 67% Claude Code co-authorship on merged PRs across 59 contributors and 35 repos, from security engineers to PMs to sales. All through systematized adoption any exec can copy. blog.trailofbits.com/2026/03/31/how…

After so many experimentations over a very long period of time, I realized something a long time ago. The model doesn’t really matter anymore, it’s the algorithm beneath it that matters even more. True AGI to me is a convergence problem. If you run a good enough capable model long enough, it can eventually achieve the same thing as “the best model”. It would just take longer because the weaker model makes more mistakes and has more learning to do. More corrective cycles, slower convergence rate, but the destination is identical. The gap between models isn’t a gap in potential, it’s a gap in time complexity. That’s what my algorithm Palimpsest is about. Making model comparison asymptotically irrelevant. I think of it like neuroplasticity. Someone with a lower IQ being able to achieve the same thing as someone with a very high IQ, is it impossible? No, because we’re humans and we’re extremely flexible. The brain doesn’t need the optimal pathway, it just needs a pathway. Synaptic reinforcement kicks in, it reroutes, it reorganizes over time toward the same endpoint. It costs more, it takes longer, but the outcome is functionally equivalent. Different substrate, same solution. That’s convergent evolution applied to intelligence. A true AGI system should make model branding ecologically obsolete. But there is a lower bound to this, a minimum threshold of baseline capacity the model needs to clear first. Below it, the system just doesn’t have the representational complexity to even encode the solution space. You can’t expect the dumbest model like GPT-2 to become Einstein, same way you can’t expect a nematode’s 302 neurons to produce abstract thought. The substrate has to be capable enough to at least begin converging. Based on my observations, that lower bound became practically reachable when o3 came out. That was the inflection point. Every foundational AI company out there is focused on throwing more and more compute at the problem, obsessing over model capacity, I’m working on the other end. I don’t think more intelligence is needed, it’s already enough. We already crossed the lowest constant bound, the threshold is behind us now.