Nilav
193 posts

Nilav
@gerceboss_21
Prev Intern @Oracle | Prev @0xcatalysis | Prev @LighthouseWeb3 | Vice President @Blocsociitr | 🏀🎸
Katılım Mayıs 2023
626 Takip Edilen170 Takipçiler
Nilav retweetledi

Wrapped up Cohort 6 with @pldevguild 🚀
This time, I worked on the @akavenetwork data-explorer indexer for a block explorer that makes events, transactions, and on-chain data actually human-readable with @gerceboss_21.
Instead of:
“Transaction 0xabc… called contract 0xdef…”
You get:
“Storage.uploadFile() by 0x…, CID=…, size=…, bucket=…, status=…”
Try it out(See readme for all instructions):
github.com/DarkLord017/ak…
Also worked on a project on @storachanetwork : an rclone-based system that lets you sync data directly from cloud providers (S3, Google Drive, GCP, etc.) to Storacha, without storing files locally with @gulshanprr.
Repo:
github.com/DarkLord017/st…
Will be adding a proper README soon.
English

Over the last 6 months, I had been researching and working on novel fee mechanisms for DeFi. If we want to onboard tradFi, we need to speak there language.
Professional hedge and liquid funds have specific accounting mechanisms like Series Accounting that they have been using for decades.
I started a discussion to standardise this method and create it as an ERC. Would love community feedback and support.
Link to PR: github.com/ethereum/ERCs/…
Link to discussion: ethereum-magicians.org/t/erc-8113-ser…
English

AI is progressing at an exponential rate and I have seen first hand how good AI auditing tools have become over the past couple of months.
Ofc these tools are no substitute for a manual review by experts but the fact that AI can work tirelessly on multiple codebases is worrisome to say the least.
This is because a single hacker may only focus on a few codebases at a time, while an AI agent trying to exploit mainnet contracts can work on multiple projects at once.
Think about it, a big project with millions in TVL and bunch of good audits is probably secure enough to not get hacked by AI. Smaller projects with a couple thousand in TVL however may have easier bugs and cheaper quality audits.
Earlier it didn't make sense for a hacker to exploit these projects because it was a lot of work just to hack a couple thousand dollars. But with AI, you can run an exploit farm, trying to hack several small projects simultaneously, and just by pure luck you may succeed in a few of them.
On its own, each hack would be small so may not be a big deal for the attacker in terms of legal threats (big firms with big layers and law enforcement agencies wont be after them with full force). Over time, a single entity with the help of AI can exploit hundreds of such projects and steal millions of user funds.
This is bad.
Anthropic@AnthropicAI
New on our Frontier Red Team blog: We tested whether AIs can exploit blockchain smart contracts. In simulated testing, AI agents found $4.6M in exploits. The research (with @MATSprogram and the Anthropic Fellows program) also developed a new benchmark: red.anthropic.com/2025/smart-con…
English

Nothing better than starting the week with really good news. Finally today I received an email by @ethereumfndn @ethereum that I've been selected as a DEVCONNECT SCHOLAR and I will be at @EFDevcon Buenos Aires, Argentina.
Awesome opportunity!
@web3mh_ @Dojo_Coding

English

After the summer internship at Oracle, I wanted to get back to strengthening my fundamentals . I am making this as a learning project and using Cursor . Checkout: [incomplete yet ] github.com/gerceboss/libp…
English

AI models and IDEs turn out to be very useful to me whenever I am starting out in any new field. They help me learn quickly and explore ideas.
But once I have a deep understanding, they just don't prove to be that great.
At that point, they're more of an execution tools for doing redundant work than being very useful as a thought partner.
English

















