Research
We're building the science behind autonomous AI agents that can reason, learn from experience, and work together to solve complex problems.
Latest Publications
OpenBlock reaches #1 on Terminal Bench with frontier agent OB-1
OpenBlock's frontier coding agent OB-1 achieved breakthrough performance on Terminal Bench, reaching the #1 position among all autonomous coding agents with exceptional accuracy in real-world software engineering tasks.
Read Full PostResearch Focus
Reinforcement Learning
Building agents that learn from real-world developer experiences and continuously improve through interaction feedback.
Multi-Agent Systems
Enabling AI agents to coordinate, share knowledge, and work together on complex tasks.
Long-Term Memory
Developing systems that remember, learn from past experiences, and improve over time.
Continuous Learning
Developing systems that adapt and improve from developer feedback to create more capable and aligned AI assistants.
Achievements
#1
OB-1 reached #1 on Terminal Bench on 09/10, the industry standard for evaluating autonomous coding agents.
View Benchmark ResultsJoin Our Research Team
We're looking for exceptional researchers and engineers who want to push the boundaries of what's possible with autonomous AI.