Hacker News Reader

This isn't Gemini, it's Poe. The blog post even admits that! I entered the exact same messages into Google AI Studio and it happily translated it to Sanskrit without any complaints about safety.

View on HN

As someone not very familiar with Poe, can you explain the difference? It looks like Poe is a front-end for multiple LLM's, and they advertise access to Gemini Pro [1].

What's the difference between the Gemini Pro they connect to, versus what you're accessing via Google AI studio? Is Poe adding additional safety checks, or is Google adding extra ones when used via third-party services? Or is Poe just malfunctioning, or not always using the model it advertises?

[1] https://poe.com/about

They apply their own filters to the responses. The message that the blogger was getting isn't one normally returned by Gemini.

Ah thank you. Then as far as I can tell this whole post should be flagged, as it's just completely inaccurate. This isn't about Gemini, it's about Poe.

You can turn off some safety filters off when using gemni pro api. I use it on my own custom chatbot

Reply on news.ycombinator.com