From Classroom to Cutting Edge: What Agentic AI and Deep Search Agents Taught Me

Mohammad Bilal — Wed, 22 Apr 2026 09:35:20 +0000

Hi! I am Mohammad Bilal, a BS Computer Science student at FAST University Peshawar, currently studying Artificial Intelligence under Dr. Bilal Jan.
For this assignment I explored two recent AI research papers. In this post I want to share what I learned and how it connects to what we study in class.
The Two Papers I Read
Paper 1: The Rise of Agentic AI — A Review of Definitions, Frameworks, and Challenges (2025)
Paper 2: A Survey of LLM-based Deep Search Agents (2026)
Paper 1 — The Rise of Agentic AI
This paper talks about how AI is no longer just chatbots that answer questions. Modern AI systems are becoming autonomous agents. This means they can plan ahead, take actions on their own, use tools like web search, and work toward long term goals without a human telling them every single step.
How Paper 1 Connects to Our AI Course
In our class we studied different types of AI agents — simple reflex, model based, goal based, and utility based agents. This paper shows that modern AI systems like ChatGPT are essentially advanced utility based agents. They look at many possible actions and pick the one with the highest expected value. This is exactly what we modeled in our rescue robot assignment.
Paper 2 — A Survey of LLM-based Deep Search Agents
This paper surveys how large language models are now being used as intelligent search agents. Instead of answering from memory only, these systems actively search the web, read multiple sources, and combine information into one accurate answer.
How Paper 2 Connects to Our AI Course
This directly connects to our search algorithms topic. In class we studied BFS, DFS, UCS, and A* to navigate a grid. These deep search agents do the same thing but at a much bigger scale. Instead of navigating a 3x3 grid they navigate millions of web pages. The heuristic in A* becomes a relevance score. The logic is identical, only the scale is different.
What I Found Using Google NotebookLM
I uploaded both papers to Google NotebookLM and asked it questions about the papers. NotebookLM helped me see the connection between both papers that I had missed during manual reading. The agentic planning in Paper 1 is actually implemented through the search based reasoning in Paper 2.
My Personal Takeaway
Everything we learn in class — agents, search algorithms, CSPs — is not just academic theory. It is the actual foundation of systems like ChatGPT and Gemini that millions use every day. Pay attention in your AI class!
Video Explanation
Watch my short video explanation here:
https://youtu.be/Qkd395Zn5QA
Conclusion
These two papers showed me how fast AI is evolving and how classroom concepts are the foundation of the most powerful technology being built today.

Posted as part of AI Assignment 1 — FAST University Peshawar
Course: Artificial Intelligence | Instructor: Dr. Bilal Jan | TA: @raqeeb_26

DEV Community: Mohammad Bilal

From Classroom to Cutting Edge: What Agentic AI and Deep Search Agents Taught Me