Poetiq claims to have achieved a new state-of-the-art (SOTA) standard in artificial intelligence, particularly within the context of the @arcprize platform, using the Gemini 3 and GPT-5.1 models. The key assertion is that more intelligence doesn’t always correlate with higher costs, challenging a common belief in the AI community. This suggests potential for more efficient AI solutions that do not necessarily demand extensive resources. The system employs iterative problem-solving combined with self-auditing capabilities. Essentially, it allows the AI to autonomously select strategies, assess if further refinement of its output is necessary, or determine if it can submit its final result. This self-improvement model means that the AI continuously enhances its solutions over time, potentially leading to more accurate and effective outcomes. For those interested, Poetiq provides access to a full analysis and the code used in their research via their website.