Poetiq, a startup founded by former Google DeepMind scientists, has emerged from stealth with $45.8 million in Seed funding and says its “AI meta system” can make frontier large language models learn faster and solve harder problems without the time and cost of conventional fine-tuning or reinforcement learning post-training. The round was co-led by FYRFLY Venture Partners and Surface Ventures, with participation from Y Combinator, 468 Capital, Operator Collective, Hico Ventures, and Neuron Venture Partners, according to the company.
Poetiq describes its product as a layer that can sit on top of multiple foundation models, pairing with systems such as OpenAI’s Chat GPT, Anthropic’s Claude, Google’s Gemini, and Meta’s Llama to generate a specialized agent for a given problem and then recursively improve that agent for accuracy and cost efficiency. The company says customers can provide a problem and a few hundred examples, rather than the much larger datasets typically required by other approaches.
The company was founded in June 2025 by co-CEOs Shumeet Baluja and Ian Fischer, both former AI researchers at Google DeepMind. Poetiq positions its approach to accelerate real-world enterprise adoption by improving reasoning performance, citing an MIT study of 300 public AI implementations published in August 2025 that found most organizations were not seeing returns from generative AI investments.
Poetiq is also tying its debut to results on the Abstraction and Reasoning Corpus (ARC AGI), created in 2019 by AI researcher François Chollet, which measures generalization in problem-solving. The company says it established a new state-of-the-art on the ARC AGI 2 semi-private evaluation set in early December, beating the prior leader, Gemini 3 Deep Think, at roughly half the cost per task by running Poetiq’s system on top of Gemini 3 Pro. It adds that after OpenAI released GPT-5.2, it incorporated the model and achieved 75% accuracy on the ARC AGI 2 public evaluation set, which it characterizes as a 16 percentage-point improvement over the previous best result.
Poetiq says it plans to use the new capital to expand deployment of its meta system for business problems where out-of-the-box LLMs have struggled to deliver reliable performance or compelling economics.
KEY QUOTES
“LLMs are impressive databases that encode a vast amount of humanity’s collective knowledge. They are simply not the best tools for deep reasoning. That’s why efforts to improve their problem-solving skills are so slow and expensive. For ARC-AGI 1 and 2, we used recursive self-improvement to produce specialized agents in a matter of hours. It demonstrates how much we can help with problems that have been too hard or too expensive for LLMs alone.”
Shumeet Baluja, co-CEO of Poetiq
“That Poetiq managed to top ARC-AGI within six months of launching is remarkable. Rather than compete against frontier models, their team of six found a way to coax more intelligence from every LLM available. Poetiq will be a must-have for companies trying to make AI work for real-world business applications.”
Philipp Stauffer, General Partner at FYRFLY Venture Partners
“Poetiq is one of the rare AI startups that doesn’t need to outcompete frontier models or pick sides. It can enhance any combination of LLMs, any native AI platform, and any AI use case. Poetiq can provide better performance at lower costs across diverse use cases by sitting on top of foundation models, and that is a unique position to be in.”
Gyan Kapur, co-Managing Partner at Surface Ventures
