About Us
Founded in January 2024 in Edmonton, Alberta, Canada, J&F AI Enterprise Inc. is building the next generation of AI: small, growable agents that discover knowledge through experience. We take the Big World Hypothesis seriously—the world is far larger than any fixed model—so we focus on systems that learn continually, verify what they encounter, and scale computation with problem difficulty, instead of relying solely on brute-force size and static pretraining.
Our direction is inspired by the research culture in Edmonton and aligns with key ideas in the Alberta Plan [2]: building intelligence through interaction, reward-driven learning, and long-horizon capability development. We believe the most valuable AI companies of the next decade will be those that make intelligence more adaptive, efficient, and trustworthy, not merely bigger.
Our approach is grounded in the Big World Hypothesis [1], acknowledging that the complexity of the real world exceeds any model's capacity, necessitating continual learning and adaptation.
- Javed, Khurram, and Richard S. Sutton. "The big world hypothesis and its ramifications for artificial intelligence." Finding the Frame: An RLC Workshop for Examining Conceptual Frameworks. 2024.
- Sutton, Richard S., Michael Bowling, and Patrick M. Pilarski. "The Alberta plan for AI research." arXiv preprint arXiv:2208.11173 (2022).
Our founding team is a group of young and talented PhDs from the Computing Science Department at the University of Alberta. As AI experts, we are determined to serve every Canadian by ensuring equal and easy access to high-quality AI products that enhance everyday life.
We translate this direction into real products and proving grounds, including AI4Culture, AI4Finance, and NewAGI – TrueIntrinsics™ Models. Together, they reflect our long-term mission: to build trustworthy, knowledge-discovery AI that works in the real world and becomes more capable over time.
We build tools that help people make sense of complex information, improve decision-making, and support long-term well-being. Our current products include AI4Culture, AI4Finance, and NewAGI – TrueIntrinsics™ Models, reflecting our work across community information, decision support under uncertainty, and next-generation adaptive agents.
Products
We are building the next generation of AI: small, growable agents that discover and verify knowledge through experience—because the real world is bigger than any fixed model. We focus on continual learning, trustworthiness, and efficient computation that scales with problem difficulty, rather than brute-force scale alone.
AI4Culture
Completed: Winter 2023 – Winter 2025
AI4Culture allowed us to understand how existing LLMs and LLM-based agents process unstructured data. We worked with news from many sources, web information (including government websites), utility announcements, transportation alerts, and legal/jurisdictional updates. We packaged our unstructured news-processing pipeline into a product that delivers timely news and summaries for people in Canada, including updates at national, city, and community levels.
We have released our official account on WeChat (gongzhonghao) that provides high-quality, reliable information for Chinese Canadians. Next, we plan to expand to additional social platforms and support more communities in Canada, in more languages.
AI4Finance
Stealth mode: Spring 2024 – Present
AI4Finance reuses our unstructured information-processing module. We are building low-frequency quantitative trading agents, based on the belief that the next generation of quant systems will be defined by their ability to digest highly unstructured information such as news and social media (e.g., X, YouTube channels, and Telegram channels), and to distill actionable signals from extremely noisy environments.
We view this as decision-making under a non-stationary environment. Therefore, we believe reinforcement learning is a natural foundation for building these low-frequency trading agents.
NewAGI – TrueIntrinsics™ Models
Stealth mode: Winter 2023 – Present
Our flagship R&D effort toward small but capable agents built for non-stationary worlds. NewAGI – TrueIntrinsics™ Models studies how an agent can learn continually, maintain durable memory, and increase compute only when tasks demand it—while remaining trustworthy and human-programmable.
Technically, we focus on RL under the Big World setting: continual learning under non-stationarity (distribution shift and changing dynamics), long-horizon credit assignment under delayed and sparse rewards, and experience reuse via selective retrieval and consolidation to reduce catastrophic forgetting. Foundation models act as structured interfaces for interaction, while the agent’s core learning is driven by reward and verification-derived signals, with adaptive compute budgeting that scales inference effort with task complexity.
Learn more at trueintrinsics.app .