Ex-OpenAI @FakePsyho says ARC-AGI is selling a "completely false narrative" about AI capabilities: "Generally, games are quite well-designed. By far the worst part about ARC-AGI is the narrative that they're trying to push. The official ARC Prize Twitter pinned post says all of the AI models are getting below 1% on it. If the benchmark is meant to be a proxy for how good models are, it's selling a completely false narrative." "The models are mentally capable of solving those problems. The…