I’ve fiddled with it recently and it’s fine but not really better than something like Gemini.

If you tell it to use metric or imperial it still mixes them, and it seems really set on forcing you into the specific styles, with little room to experiment, like if you want a slightly more amber-leaning WCIPA (which is a very common deviation from the BJCP guidelines) it will push back because it wants you right in the middle of the style range, the XML it spat out in my test didn’t exactly match the summary it gave either. That was a week ago so maybe they’ve been tuning it.

I just feel like AI recipe generation will lead beginners astray because they dont have the experience to know whether it’s correct or not. The AI will always say something that it comes up with is amazing and special regardless of errors, that’s how they are biased, they will only correct those flaws when you point them out.

The only thing I can see AI brewing assistants being good at currently is to say things like:

• “how can I adjust this existing recipe to a full volume mash/no-sparge brew without changing the end result too much”
• “how do I adapt my all-in-one brewer process to a 2V or 3V setup, take me through the steps”
• “how can I approximate this all grain recipe with partial mash brewing”
• “I like this recipe but want it to be more floral, what do I change”