Comparing BARD to GPT-4 (and then 3.5 legacy) on a mid-level task: make us a rate-limiting queue.
First prompt:
Write me a queue in typescript that can take a number of thunks, and execute them while making sure that only n thunks are executing at any given point in time.
Off the bat Bard's code is more complex, but has less descriptive variable names for the private members. Reminds me of myself a good few years ago - if no one sees them, why name them well?
Bard's code is also invalid - unshift can return undefined.
Hrishi
@hrishioa
In SF in March - Building artificially intelligent bridges at Southbridge, prev-CTO Greywing (YC W21). Chop wood carry water.