What happens when the entity committing a fraudulent action is an autonomous agent rather than a person? Benjamin Nweke examines the new explainability challenges emerging in agentic AI fraud detection. towardsdatascience.com/what-…

Partha Sarkar put a two-tier cache in front of his agent: a semantic cache for near-identical questions, a retrieval cache that reuses stored context for reworded ones. towardsdatascience.com/shipa…

"So here's the pivot. Stop training a network to know what to do. Train one to know what happens next." Anubhab Banerjee presents a beginner-friendly guide to building a world model in Python. towardsdatascience.com/how-t…

In his new explainer on GRPO, @BenjaminNweke11 unpacks the mechanics behind local reasoning and explains why the reward function matters as much as the model. towardsdatascience.com/how-g…

How do ML systems process language and generate responses? @Nikhil_on_ds unpacks the journey from words to vectors in his recent article. towardsdatascience.com/from-…

"If you look inside a trained neural network hoping to find individual neurons that cleanly represent individual concepts, one neuron for 'is this a dog,' one for 'is this red,' you mostly don't find that." Utkarsh Mangal shares insights based on his work to recreate Anthropic's superposition paper from scratch using NumPy. towardsdatascience.com/i-tra…

Watermarking is moving beyond images and video into the words AI systems generate. Chien Vu Minh examines how these techniques work in practice and which approaches hold up when text is copied, edited, or paraphrased. towardsdatascience.com/text-…

The problem with coding agents is not always what they say, but how they present it. Eivind Kjosbakken shares practical ways to make terminal-based communication clearer and more effective. towardsdatascience.com/how-t…

Dimensionality reduction can make complex classification problems easier to understand and visualize. Carolina Bento demonstrates how Linear Discriminant Analysis works using a real-estate dataset. towardsdatascience.com/linea…

How much can you actually get out of GPT-6 Astra? Eivind Kjosbakken puts the new model through extensive use and shares the approaches that helped him maximize its effectiveness. towardsdatascience.com/how-t…

Fraud detection is entering a new phase as AI agents begin making transactions on behalf of people. Benjamin Nweke explores why traditional explainability methods like SHAP may not fully capture agentic fraud. towardsdatascience.com/what-…

Are we witnessing a new type of model eclipsing LLMs in real time? It might be a bit too early to tell, but if you'd like to learn about Jev, a new System One model, don't miss @taupirho's accessible introduction. towardsdatascience.com/an-in…

A source says a model reduced error by 20%. The draft says accuracy improved by 20%. Same number, different claim. Benjamin Nweke built EvidenceDesk with Codex to catch that kind of drift before a technical article goes out. towardsdatascience.com/shipa…

Follow along @taupirho's latest tutorial to learn how to benchmark the impact of fewer, larger files across three SQL workloads. towardsdatascience.com/i-com…

In an insightful analysis, Yu Dong revisits a timely topic: the shape and extent of AI's impact on data science roles. towardsdatascience.com/ai-ma…

"This has never sat right with me: a bounded yes/no or routing decision was often passed through the same autoregressive decoder used to generate a paragraph." Lambert Leong @leonglambert digs into the architectural constraints that are steering the field towards "decision" models like Jev. towardsdatascience.com/when-…

Getting started in designing safe and trustworthy agentic systems? @Thuwarakesh presents a comprehensive guide to the architectural guardrails you need to be fluent in. towardsdatascience.com/how-t…

We're thrilled to welcome Vasileios Vonikakis to TDS! You should explore his debut article, which unpacks the process of building fair evaluation sets when dealing with imbalanced data. towardsdatascience.com/build…