Bronto

the challenges I run into, the mistakes, and the solutions I discover along the way. More soon: t.me/web3securityre…

None of these projects are close to production yet, but I genuinely believe they solve real problems. So instead of waiting until everything is "ready," I'm going to start sharing the journey.

Spent this weekend at the Hack Nation (linkedin.com/company/hack-n…) hackathon. Out of 6 tracks, our team chose to build a voice AI assistant using ElevenLabs. For ~20 hours of hacking, I'm pretty happy with what we managed to ship.

@VittoStack Hello, Vitto! Requesting to join.

4 days ago we launched Jailbroken, a PRIVATE Discord community to learn AI red teaming and safety. Since then: - Over 250 security researchers joined - Top resources have been collected - People shared countless techniques and discoveries Today, we've secured over 100B in FREE AI tokens for all the members. If you want to join, drop a comment.

6/ Deep dives into the architectural decisions will be posted on Noobing Security Research. Follow along 👇 t.me/web3securityre…

5/ I'm solving two problems at once: → building a system that makes my own work easier → figuring out, hands-on, what it actually takes to build a secure agent

"What's that in your agent? A hole." I've reached an obvious conclusion: almost nobody thinks about how secure their agents actually are. So I'm going to show - in practice - how to build a more secure system. 🧵

@zacodil That's just crazy! It turns out that this kind of L2 network is less fault-tolerant than traditional financial systems

Base was down for almost two hours today, and the reason is the whole story: one bad block, and nobody else to keep the chain alive. A single consensus bug sequenced one invalid block, right after block 47806542, and every block after it just stopped. Base runs a single sequencer, so when that one machine choked, the entire network choked with it. no backup producer, no other validators to route around the problem. The exit was worse. the chain couldn't heal itself: node operators across the ecosystem had to manually restart before syncing recovered. and for those two hours there were no blocks, which means no trades, no liquidations, no withdrawals. if you wanted out, you couldn't, because leaving is also a transaction. And this isn't the first time. a sequencer handoff failure froze Base for 33 minutes last August, today it was nearly two hours. that's the part the "L2 is a blockchain" pitch skips: a chain with one sequencer isn't decentralized infrastructure that happens to be fast, it's one server with great branding, always one bad block away from a full stop.

кто то навайбил площадку для сбора информации о баг баунти скаме bugbountyscam.com. Если вас когда-то загостили, вам не заплатили или что-то в этом роде, то можно оставить отзыв

5/ пока мы не увидим примеров мощных взломов, я думаю в этом году уже будет такая история 3. Совершенно точно стоит заниматься безопасностью агентских систем 4. Ждите на канале теорию по векторам атак и разборы взломов t.me/web3securityre…

4/ 1. Пока что потребности специалистов в большинстве своем закрываются существующими подписками со встроенными инструментами, даже при высоком уровне интеграции в сложные задачи 2. О безопасности будут думать не только лишь все до тех пор,

1/ Слушал сегодня доклады на AI конференции aiconference.communitysprints.ru Спецы рассказывали свои кейсы использования агентов в работе, два часа докладов. Уровень от использования клода на полную, до вполне себе агентских систем в связке с навайбленными инструментами Тезисы такие: