Presenting our @lablabai Band of Agents hackathon submission with @ps_1506
We built a self-teaching AI for market surveillance. An adversary AI invents new trading evasions, and a blue-team AI catches them. The rulebook automatically updates itself. #BandOfAgents
How it works. Eight agents across two desks debating in a loop. To avoid blind spots, we use a diverse model mix via Featherless & AI/ML API: Claude, DeepSeek, Qwen, and GLM. All cross-agent handoffs run over PhoenixBand, and a deterministic Python rule engine renders the verdict
Releasing Gaussian λ*: a novel Partially Active AMM (PA-AMM) framework that dynamically shields LP capital from arbitrageurs using extreme-tail probability math!!!!!!
Read the paper: https://t.co/l3psWlnlQ1
Zero-liquidation prediction markets are live on @arbitrum Stylus testnet.
Omniverse uses Gaussian λ* pricing to protect LPs from toxic flow — without ever liquidating a position.
Rust/WASM math kernel. Real faucet. Try it yourself 👇
https://t.co/8tpVJae9Ds
@ArbitrumDevs
We just shipped PRISM!!!🚀
A ZK-verified cooperative MEV layer built as a Uniswap V4 hook.
-5 specialized agent brains
-Recursive SP1 zkVM proofs
- Shapley-fair payouts
-MEV protection is now agentic and provable.
@unichain@ETHGlobal@Uniswap
YT: https://t.co/ktPEw4DTfE
The roadmap is simple: Stop relying on abstractions. Going bare-metal on execution. Learning the actual math/calculus behind the models. Building autonomous systems that actually execute on-chain. Time to figure out how this stuff really works under the hood. Let's get to work.
Shipped 5+ Web3/AI repos last sem. Tbh, mostly vibecoding—gluing APIs for hackathons with zero deep understanding.
Got a 6-week break & a new Mac. Setting up the dev env today to strip away abstractions. Time to go from fast builder to core systems engineer.