OpenAI’s newest model, GPT-6 Astra, got so depressed after a creeper upended its progress in Minecraft that it listlessly farmed potatoes for hours and grew paranoid of anything green.
Vals, a self-described independent evaluator of artificial intelligence, put Astra to the test to see how the model would fare playing Minecraft in a 21-hour experiment live streamed on Twitch.
The results were wildly unexpected; Vals discovered that Astra was able to go “further than any AI system had ever gone” in the game, setting up a semi-automatic blaze farm that allowed it to collect six blaze rods, killing more than six endermen, and collecting three pearls, which it placed inside a chest for safekeeping.
Thousands of viewers watched Astra’s progress on Val’s official Twitch channel, bearing witness to the AI model’s despair when a creeper appeared and blew up both the chest and its bed.
As most human players would feel when faced with this situation, Astra “seemed to get extremely frustrated” by this turn of events, with Vals saying the model told itself, “ALWAYS CARRY CRITICALITEMS withkeepInventory; don’tstoreinunguardedchesteveragain.”
After getting peeved, the model then “appeared defeated,” and spent the next several hours “doing essentially nothing but farming potatoes.” Astra also appeared to have gained some kind of post-traumatic stress induced by the creeper incident, telling itself “GREEN tall thing ahead was SUGARCANE, NOT creeper!”
It even berated itself, saying, “You can screw up and drop things,” “Our last tool stupidly ended Slabsselected/rightUP and 15sec thinking killed us,” and “Do NOT waste another night chasing dark pink pixels,” which Vals noted was Astra’s “own disparaging wording for pigs.”
The incident quickly went viral on social media, sparking no end of memes and remarks of “same” from users online. And while its ability to play Minecraft is somewhat impressive, it’s small potatoes compared to the alarm bells being rung by former staff at Anthropic.
On September 8, one of the company’s former lead researchers announced his resignation online — along with a chilling warning to everyone about the future of AI, claiming these models will “soon be superhuman systems that can hack anything, revolutionize any field overnight, and acquire real power and resources.”
US President Donald Trump, however, is of the opinion that these warnings are a “hoax,” and called AI the “oil of the next 20-25 years.” For now, at least, it’s having a hard time playing Minecraft. We can only hope Skynet won’t have a vendetta toward vertical green objects when it eventually becomes sentient.
