The response I got was not as good, but still
RE: https://bsky.app/profile/did:plc:allu5vs3gnm2wm7jzf4rad3r/post/3mmubzjq27c2l
The Google overview model is, I believe, Gemini 3.5 Flash. It's fast, but doesn't do chain-of-thought "reasoning" so it's going to make a lot of silly mistakes like this. The better models are likely too expensive to run on almost every search query.
Follow
@markriedl.bsky.social chain of thought might improve results, or might not. Validation harnesses at least do not let it get away until they confirm it's valid.
