ilemi

ilemi

20,031 Twitter followers

Follow

agent @herd_eco | advisor @archetypevc | ex-@dune 👇Start your data journey (ethereum, solana, bitcoin)

Posts

Just landed in Cannes - I will be here for the next week :) hmu to talk about how we help institutions and curators with continuous due diligence of vaults/tokens, or about how to create agents with herd that can execute safely across all of defi (not with unstable skills lol). twitter.com/andrewhong5297/sta...

claude is still updating this every 30 minutes, with clear timelines It now contains details on attacker wallet funding, all 5 attacker wallets activity, as well as the team multisig response. funds have been consolidated into one wallet now, but have not moved in a while twitter.com/andrewhong5297/sta...

TOMORROW. 11:20AM ET. @andrewhong5297, Co-Founder of @herd_eco

Every token should have its minting system transparently monitored. Here are the minters for USDC on Ethereum, for example, many contracts/multisigs/eoas are involved in nested levels. However, they have a max mint amount assigned in the role mapping for extra protection.

gm, looks like the USR attacker has three wallets now and is working on dumping wstUSR still

having claude update a github gist with everything about the USR hack and the attacker + team multisig activity. two hacker wallets to track right now gist.github.com/andrewhong5297...…

Just being able to docker it alone was enough of a benefit for me lol Plus the better setup ux and native honcho integration

Is Hermes really better than Openclaw ??

First time looking at the frame tx spec, feels like a code smell that there are so many different configurations/fail cases. I personally liked the dx of the tempo tx a lot. Takes inspo from solana model which does it well with instructions. Execution needs to be easier to reason about, not more complex. twitter.com/andrewhong5297/sta...

Think bankrs iterations here are interesting. We started with pure “abi -> tx” back in 2024 but found that the variance in tx outcomes over 100 runs was too high to trust with large sums of money (especially for more complex functions than deposit/withdraw). We’ve ended up giving the agent a transaction building language it could self-define and iterate over, and have found great success here. But its still interesting to think about when you need a skill vs a language/code, and what the minimal version of each is. twitter.com/andrewhong5297/sta...

Rebalancing for risk is an agent strategy we're testing with a few teams now. The idea of an agent tracking vault functions/roles by counterparty, liquidity, and operational risk, and setting up guardrails against them. It's a more qualitative and dynamic analysis than just yield optimization, which is what agents excel at. imo the current agent "optimizations" are really just if/else formulas, not truly agents used to their full potential.

The idea that user run agents will continuously rebalance across vaults is so compelling, but the math isn't great for highly active rebalancing Average rebalance tx fee: $0.001 Rebalance rate: every 10 seconds Annual rebalance cost: -$3,154 Deposit size: $100k Baseline Yield: x.com/lui1of1/status…

Loading..