
Introducing: Arenita
The petít artificial (and human) intelligence arena.
Our take on evaluating Human and Artificial intelligence
in a single loop
It’s been a while since we launched something new. If you’ve been following our updates, you’ll know that almost all the demos and processes we had planned are already live and validated on their own.
Because the incubation process requires us to focus on a single product or technology, we saw this as a great opportunity to develop something new. It is also a chance to experiment with different stacks and architectures, leveraging what we’ve learned from our original demos and our recent experiences in other industries.
What’s Arenita in this context?
We are creating a very straightforward platform for humans to test and evaluate LLM models. It works like any other double-blind arena, but on a smaller scale, as that isn’t its only purpose.
It works like this: you ask the bots and pick the best answer, with the possibility to leave notes and observations. These chats and conversations essentially become open data that we share back with providers and the community to keep training and improving the models we all use.
What sets this platform apart is that the account you create to access it becomes a profile that qualifies and quantifies your participation through a flexible points scheme. Top participants will be invited to submit their CVs and take additional tests to begin building a psychometric profile. This way, anyone can become part of the worker pool we discussed in the Opportunities Navigator.
The basic service is free from the start, and while the advanced features will also be free, they are currently available by invitation only.
What does ”arenita” mean?
The name alone gives away the initial focus of this exercise: the Spanish language. Let me explain this for those whose first language isn’t Latin-derived.
In that linguistic family, “arena” means sand; it is almost the same word the Romans used to describe the pits of the Colosseum (harēna), which were literally covered with harēna. English already had a word for “arena”-which was “sand”- so when English borrowed the word arena, it went straight to mean “venue,” which makes perfect sense.
Now, with the rise of “the arena” in the context of AI, the meaning has come full circle back to a place for competition and duels. That is the context of our platform as well. In Spanish, you can add an -ito or -ita to almost any word to mean “a little” of something. In our case, a little arena.
In this region, “arena” is also associated with “venue”; and “sandbox” is also a common concept (cajón de arena). That last one also aligns well with what we are building here.
What’s the status of Arenita right now?
Arenita is now available for early access. You can create your account for free right now and start earning participation points to get invited to advanced features.
For basic access—the raw arena part—we currently offer open access to several local models. These may be slower or less reliable, which is actually good news for you: errors provide more “content” to evaluate. You can also report glitches regarding the platform or its processes in the observation fields for each answer.
We are currently working on grants and partnerships to offer the latest models from all providers and to move the advanced features to public access. Since we plan to use Arenita for our own selection processes, we aren’t in a rush to push just another step in the development cycle. It may be a better plan to launch fully once we have validated the entire process; having the full picture will allow us to simplify everything as much as possible.




No comment