6 min
The Real Challenge in Search Agents Is No Longer Searching More
Perplexity Research's new article shows why the next phase of search-augmented LLMs is about reward design, tool-use efficiency, and cost-aware agent behavior.
/blog
Short essays about AI systems, data science, engineering habits, and the career bits that deserve more than a LinkedIn post.
6 min
Perplexity Research's new article shows why the next phase of search-augmented LLMs is about reward design, tool-use efficiency, and cost-aware agent behavior.
5 min
Claude has become incredibly capable, but under heavy real-world use, token and credit burn starts turning into a workflow problem.
4 min
Webcam-based eye tracking is accessible, but its accuracy can change dramatically. I built a browser-based system to measure that reliability before using gaze data downstream.
6 min
EASIEST explores a simple but ambitious idea: can a normal webcam collect useful gaze evidence for ASD screening support inside a full web application?
5 min
Applied Bayesian analysis is not only about estimating effects. It is about modeling uncertainty, checking predictions, and comparing explanations responsibly.
5 min
Some scientific models can simulate reality better than they can write down a likelihood. SBI is the bridge between those models and Bayesian inference.