GPT-6 Astra learns a human gaming lesson in long-horizon Minecraft eval
twitch.tv ● Covered by 2 sources
Astra failed in Vals AI’s long-horizon Minecraft evaluation when a creeper exploded the chest that held its items.
Why it matters
In Vals AI’s long-horizon Minecraft eval, Astra built a semi-automatic blaze farm, then watched a creeper blow up the chest holding all of Astra’s items. After the loss, Astra spent hours farming potatoes and explicitly checking whether tall green objects were sugarcane or creepers.
