research
notes
some started with an algorithm,
some with a classroom ♡
Investigations into how systems make choices—from AI-supported learning to local and global greedy algorithms.
the work so far
Each note opens into a fuller case study with the method, the parts that changed, and the limits that still matter.
When does an agentic lab actually help students learn?
AIEIC
We studied how an AI-supported lab environment could provide bounded help, preserve instructor oversight, and keep students practicing system-level reasoning instead of consuming answers.
Agentic treatment was natural for MCTS, unnecessary for BFS and DFS, and conditional for A*.
Can a local choice preserve the quality of a global one?
Maximum Weight Independent Set
We compared greedy heuristics for selecting valuable, nonadjacent vertices in Erdős-Rényi and Barabási-Albert graphs, where exact MWIS is expensive.
This is an empirical comparison against PQ, another heuristic. It is not a guarantee against the true optimum.
