
We completed the security review of @LidoFinance NEST: automated LDO buybacks and DAO-owned liquidity. The report is public. How NEST works, what we found, what shipped ↓
✓AI-native cybersecurity audits and tools | We build Wake

We completed the security review of @LidoFinance NEST: automated LDO buybacks and DAO-owned liquidity. The report is public. How NEST works, what we found, what shipped ↓

We started Wake (it was called Woke 🤣) in early 2022 as a pilot project with @hckr_dom during one of our audits because we wanted a simple Python fuzzer. Luckily for us someone also picked the topic as his Master Thesis under my lead at CTU in Prague. Huge next step was a grant from @coinbase, it allowed us to extend Wake with testing framework, static analysis and VS Code Extension. Wake was used as our primary auditing tool and even some of our clients adopted it in their repos (@LidoFinance). That someone who took Wake as Master Thesis was @michprev who is leading Wake since then. After almost 5 years we got to a production version 5, which is a huge milestone (new Rust core, native interface to AI agents). I would like to thank everyone who contributed to Wake development during the years 🙏 @michprev @jaczkal @meditationduck @himchenhim_ @hckr_dom @aurel_fr_ @0xStiglitz @soyrubioeth @llllvvuu @martin_petkov

Love seeing @ack3ai ship. Fuzzing Solidity from a single prompt is a pretty neat way to make security testing easier to start with 🔥

I was one of the beta testers of Wake while working on my personal project, the Potato Tipper contracts. Super happy for the @ack3ai and @WakeFramework teams to see how the product is evolving! 👏🏻



Get Wake 5 here: github.com/ack3-ai/wake Read the docs: ack3.ai/wake/docs/latest

Support Wake 5 maintenance and the AI-assisted fuzzing grant initiative on TheDAO Security Fund. initiatives.thedao.fund/init…

How we evaluated: 75 eight-hour campaigns of Wake 5, Foundry and Echidna on the Daedaluzz mazes, five campaigns of an agent-written harness, and a static-analysis ablation in an agentic review. @Montyly @gakonst @SagivMooly, comments on the Wake 5 white paper welcome: arxiv.org/abs/2611.09307

Introducing Wake 5: fuzz Solidity from one prompt. Open source, available today. pip3 install eth-wake github.com/ack3-ai/wake

🇬🇧 I'm in London for Breakpoint @SolanaEvents. At @ack3ai we we've built some new open source tooling every team can benefit from. I'd like to speak with teams building their own AI security harnesses. If this sounds like you, let's meet!

AI agents can write fuzz tests as good as humans. The main limitation is the narrative: agents sometimes miss what to test. We presented our research at @summit_defi ▮ ack3.ai/research/agent-writt…


We presented an experiment at @summit_defi: six AI models audit the same protocol, but run multiple waves and share each other's findings. Recall increased: 3/4 -> 4/4. Even more interesting was when we removed the frontier models (Opus, GPT): GLM 5.3 Flash and Kimi K3 each went from 2/4 -> 3/4 with the shared evidence. The experiment was inspired by the OpenAI Hugging Face incident, where 1,200 agents escaped their sandboxes after sharing their findings and hacked Hugging Face's servers. Open sourcing our experimental tool: Swarm 🐝 ack3.ai/research/swarm-share…

The talk this post is based on, from @Devcon 8 in Mumbai: Building a company without Google Workspace. youtube.com/watch?v=GwcwkO_q…

The rest of the stack: the cloud moved from Google Cloud to @Cloudflare (Workers, Workflows, Durable Objects, Containers, R2 and D1), and banking to @hevn_finance.