This is awesome! Claude does have an incognito mode so I tested your queries there and made a copy of your doc with its responses along with my ratings included:
Cracking work! Love the initiative of you extending this. Interesting to hear that Claude is EA-pilled, but that its animal welfare opinions are still more vibes-based than taking the same, rigourous stance
Deeply appreciate this Kristof. Interesting that at a broad level (what are the best charities, how to help more people), it cites credible and evidence-based resources.
Then when discussing animals and Africa—i.e. more long-tail keywords—it does not.
There is probably low-hanging opportunity here for charities to write up more indexable FAQs and blogposts that match the language a user would use when asking an LLM a question (or even Google).
This is awesome! Claude does have an incognito mode so I tested your queries there and made a copy of your doc with its responses along with my ratings included:
https://docs.google.com/spreadsheets/d/1Qlq9hTLQ4U7BPdMNN7Rsxxo4A8GLDv0tfYjll4MKkW4/edit?gid=0#gid=0
It did a bit better than ChatGPT and Gemini but still below Grok. I’m pretty impressed with how “EA-pilled” it was overall.
Cracking work! Love the initiative of you extending this. Interesting to hear that Claude is EA-pilled, but that its animal welfare opinions are still more vibes-based than taking the same, rigourous stance
Deeply appreciate this Kristof. Interesting that at a broad level (what are the best charities, how to help more people), it cites credible and evidence-based resources.
Then when discussing animals and Africa—i.e. more long-tail keywords—it does not.
There is probably low-hanging opportunity here for charities to write up more indexable FAQs and blogposts that match the language a user would use when asking an LLM a question (or even Google).