Almanak launches its Agentic DeFi strategy platform, allowing users to describe trading ideas in natural language, which are converted into Python code by AI agents, backtested, simulated, and deployed on-chain.

The piece highlights two critical risk points: boundaries when agents touch real funds and the credibility of backtests, warning that strong backtests don’t guarantee future profits due to issues like overfitting.

The overall takeaway is that lowering barriers to AI-driven quant strategies increases access but does not reduce risk; discipline and risk management remain essential.

Backers include Delphi Labs and HashKey Capital, signaling strong investor interest.

Almanak emphasizes a non-custodial, permissioned approach where strategies deploy to the user’s Safe wallet, with Zodiac Role restricting agent permissions to keep private keys with the user and reduce unauthorized actions.

The platform distills three defensive designs for safe AI agent use: least privilege permissions, shadow-run first (mainnet fork/testing before live), and a rollback/kill switch to halt actions quickly.

Caution that AI-generated strategies can be fast, but real-world factors like gas costs, slippage, and MEV can erode returns, widening the gap between backtest results and actual performance.

At launch, seven chains are supported (Ethereum, Arbitrum, Base, BNB Chain, Polygon, Optimism, Avalanche) with strategy types including liquidity provision, yield optimization, delta-neutral, and tokenized assets.