Can a gemini session compete?

I created a particular session with gemini-flash-1.5.
As a result, once loaded the session, gemini behave in, let’s say, a very particular way.
How could I make this session compete?
I mean, I could use any generic app that loads the session and then starts the inference as a continuation of that session, but the app itself won’t have anything particular.
P.S. the session is now long around 300K tokens and the result is incredible.

Can it? Certainly. Anything can compete.

But I think you need to ask what your chances are and consider what the goal of the competition is and what your competitors might submit.

The competition sure seems like it is trying to encourage advanced applications. Not just advanced prompts. The goal seems like it is trying to push the boundary of how people will use Generative AI.

So an input that is added to a single (long) prompt to get excellent results may be fine, but is it a good User Experience, no matter if it is a good answer?

If you look at the judging criteria in the rules, there are 15/65 points for “creativity” in your methods and UX. There is another 10 points about executing and engineering it in a good way. I’m not sure a chat bot with a single prompt behind it can get many of those 25 points. There are 10 points for “remarkability” - ie: will someone who is familiar with GenAI and prompting be impressed by this? Will someone who isn’t?

It is also clear that many of the prizes are seeking well-rounded applications. There is no prize, for example, for best prompt. There are prizes for “redefining what is possible” and for a variety of UX technologies.

You are welcome to do as you wish, of course. And I look forward to hearing about it after the competition! But this competition is trying to kick things up to the next level.

1 Like

It’s not an advanced prompt but a big session of around 200K tokens.
This lead to very interesting chracteristics I am not going to disclose for now.
Also because I’m thinking of polishing even further by using a bit of training.

400k tokens as of now and gemini is struggling already… takes more than a minute to answer… sometimes two. but the result is quite incredible.