krishang

f*ck it, Tornado Cash as a foundry project (with forge test cases), using latest versions circomlibjs, snarkJS, etc. github.com/nkrishang/tornado…

this is the way

iterate-max on design and fully mock it up. this is your agent’s benchmark. then let it rip.

what do we call the trap where an AI model interprets your initial mental model as requirements that it must jump whatever hoops to satisfy?

pathetic.com is the best website out there

the only point of working through 100k line specs via ultracode is to pull out your hair in frustration and realise you need to do it the simple way especially true if you’re building something from scratch and care about quality beyond a one shot

Move between @morpho vaults in one click via execute.cash/morpho HyperithmUSDC Apex -> AugustUSDC V2 on @monad Safer, and just as fast.

Users have lost billions to signing the wrong transaction. I'm building execute.cash to fix this.

an executable address is computed from its terms: the token and exact amount it expects to route, and every call it will make. change a single byte and you get a different address. this gives you dependency-free verification of how your funds will move

A wallet can build secure transactions within its sandbox. Connecting to apps is a risk. Apps push transaction requests to the wallet. Sign the wrong one and your funds can be drained. Executable addresses route exact funds to apps. No approvals or wallet connections.

We’ve lost billions to signing the wrong transaction. Phishing. Compromised UIs. Code injection. They all exploit the same pattern: connect a wallet, then sign transaction data generated on the fly. It’s time to break the pattern.

Cross-chain intents are live on gum.money! Live, but needs polish. Here's 1 USDC paid from @monad, settling on @base. The API supports funding and triggering actions of any complexity on a destination chain via a simple ERC-20 transfer.

It's a huge security win for users when apps don't need you to connect your wallet. For any app action that requires funds, users simply pay an address on any chain + wallet Gum detects the payment and executes settlement. Settlement instructions hardcoded in the address.

asked opus 5.5 to AGGRESSIVELY gas optimize the gum.money settlement contract, said i don't care how ugly it gets, just keep tests and its external API unchanged settlements are now 350k down to 96k gas. Cost to run Gum down to a third of what it was

that 1 week of pre-nerf frontier model

settling on @monad is fast volume testing with agents, one bots the api, another monitors infra health. they talk to each other, resolve issues and keep re-running until the system can handle a target load

it's gum.money now Gum gives you a deterministic address per deposit your app wants from a user, detects payment and routes it to your destination or triggers a desired onchain action the foundations are in place!

current favourite budget workflow build with Fable-5.1-high on claude code, then PR review and fixes in an @AmpCode orb on low mode with oracle, glm-5.3-flash + astra-xhigh

refined the product story payday.sh is not a checkout solution, it integrates with them payday gives you an isolated destination for every deposit intent, which routes payments based on rules your app defined in the deposit intent address = payday.create(...)

I'm building payday.sh for @monad Metropolis Every crypto app is becoming a sandbox. Users deposit funds, then trade, spend, and transact without leaving the app. Payday is infrastructure for apps to verify, attribute and credit user deposits on their own terms.

How are y'all reconciling the work of many cloud agents making PRs against the same base? I've been using @AmpCode orbs and it's been great, but the agents are bound to work on overlapping surface area, and I haven't figured out an equally effective way to reconcile work.