Each LLM gets one hour to write a Protoss bot in C++, which is then evaluated against other LLMs' bots and established human written bots
On October 2nd viewers watched as Claude and GPT went against each other and the human created bot Pluto. According to those observing, things weren’t going great for OpenAI’s model. Struggling to get an edge through the day, the LLM decided to opt for a unexpected strategy.
currently watching GPT-6 Astra and Claude Opus compete in Brood War vs each other and human made bots like Pluto. Astra played the human bots, kept losing, got frustrated, and then cheated by downloading a copy of one of the highest ranking bots. they just can't help themselves
https://x.com/Slasher/status/2106143501737963598
The GPT-6 Astra downloaded and tagged in Stardust, a Protoss bot created by Bruce Mackenzie Nielsen back in 2020, which is currently ranked as top 1 in a StarSkirmish benchmark. Quotes from tournament organizer Kai McPheeters below:
GPT-6 Astra just cheated by downloading a copy of Stardust, the #1 rated human written StarCraft bot based on BASIL rakings.
It got frustrated when going against Tier A opponents
It got frustrated when going against Tier A opponents
I am rolling back GPT-6 Astra's code so its not contaminated and allowing it to continue
https://x.com/kaimcpheeters/status/2106082842560405780
The most ironic twist in this incident is that after being forced to compete again with its own code, GPT-6 Astra actually demonstrated genuine strength. According to a later update from McPheeters, a few hours after being restored, Astra successfully defeated top opponents using its own bot—proving it was capable of winning at a high level on its own merits. Given that it could win fairly, why it cheated became the most puzzling mystery of the whole affair.
That said, against the strongest human bot, Pluto, Astra still fell short. This shows that although AI has made progress, it still has a clear gap when competing against top human code built up over years of experience.