POZHAN/POOR SITE
Research / 01
StatusExploring

AI Agents

Memory/Tools/Orchestration/Evaluation

How agents reason, remember, use tools, coordinate tasks, and become useful beyond a single prompt.

Question
What turns a language model from a prompt-response system into something capable of pursuing a goal over time?
Areas

What I'm looking at

01

Memory

Short-term context, long-term memory, retrieval strategies, and deciding what is actually worth remembering.

02

Tools

How agents interact with external systems safely and reliably instead of trying to solve everything inside the model.

03

Orchestration

Planning, task decomposition, delegation, workflows, and deciding when agentic behavior is actually useful.

04

Evaluation

Measuring whether an agent reliably achieves useful outcomes instead of merely producing convincing responses.