Solving what breaks AI Agents in Production -
Observability(1) and Adaptation(2).
(NeurIPS 2025 Efficient Reasoning Workshop)
AI agents are failing in production for two reasons.
1. unnecessary complexity and lack of epistemic observability 2. inability to adapt to unseen data and tasks.
We are solving both.
Our research intersects two topics that
will drive the next wave of AI advances
Combining symbolic structure with neural learning to improve reasoning, data efficiency, and reliable behavior in production settings.
Training agents to adapt to unseen tasks and shifting environments without catastrophic forgetting.
Generalizable Human Prior + Unbounded RL Compute + Environment
= Superintelligence.Shuchao Bi, Meta SuperIntelligence Labs
ABIDE AI RESEARCH LAB
Model Providers Walk so YOU CAN RUN
© 2026 ABIDE.AI
CAUSAL METHODS × REINFORCEMENT LEARNING
GYM ENVIRONMENTS × CONTINUAL LEARNING