Testing racial predictions with BISG

I Spent an Hour on a Data Preprocessing Task Before Asking Gemini

A data scientist shares a hands-on experience spending an hour manually solving a Pandas data preprocessing task — extracting a probability value by matching a category ID across list columns — before asking Gemini to do the same. Gemini solved it in seconds, but its first solution used the non-vectorized apply function. Only because the author understood Pandas well enough did they recognize the suboptimal approach and prompt Gemini for a vectorized alternative using explode and regex. The takeaway: LLMs dramatically boost productivity, but domain knowledge is still essential to evaluate and improve AI-generated code.

{golem} 1.0.0 is here

The {golem} R package has reached version 1.0.0, marking a stable API milestone for building production-grade Shiny apps as R packages. Key additions include agent skills support (Claude Code / AGENTS.md integration) so AI coding assistants understand golem conventions, reworked multi-stage Dockerfiles with production mode enabled by default, new CI helpers for GitHub Actions and GitLab CI, modernized console output via {cli}, and functional JavaScript input/output bindings out of the box. The release also includes breaking changes to path arguments, config functions, and add_/use_ helpers that existing users should review before upgrading.

How to Put your Course Book Online

A step-by-step guide to publishing course materials as a free online book using Quarto and GitHub Pages. Covers creating a Quarto book project, choosing an appropriate Creative Commons license (CC BY-NC-4.0), writing a useful README, setting up a GitHub repository, and configuring GitHub Actions to automatically render and deploy the book whenever changes are pushed. Also explains using a DESCRIPTION file to track R package dependencies for the CI/CD workflow.

I am Leaking My EX-Boyfriend’s Chats to Show How Not to Text.

Một nhà phân tích dữ liệu thu thập hàng nghìn cuộc thảo luận về mối quan hệ trên Reddit bằng API, sau đó sử dụng text mining và sentiment analysis để phát hiện các mẫu hành vi nhắn tin và giao tiếp trong tình yêu. Kết quả cho thấy sự nhất quán, nỗ lực chung, trung thực cảm xúc và giao tiếp rõ ràng mới là yếu tố quan trọng, thay vì các lời khuyên phổ biến như "phớt lờ 3 tiếng" hay "làm khó dễ".

Một lập trình viên nên đọc bài này để hiểu cách xây dựng hệ thống phân tích văn bản hiệu quả từ dữ liệu thực tế về giao tiếp tình cảm, giúp phát triển các giải pháp AI thông minh hơn trong việc dự đoán và cải thiện chất lượng mối quan hệ thông qua dữ liệu.

Building an AI-powered commercial intelligence dashboard

A development team built a commercial intelligence dashboard for a multi-brand marketplace client who wanted AI used throughout the project. The post walks through five phases: event tracking, computing user fingerprints (where deterministic algorithms beat LLMs), synthetic data generation, AI-powered brand briefing generation using an agentic workflow with RubyLLM tools, and dashboard implementation. Key lessons include recognizing where AI adds value versus where it doesn't — fingerprint computation needed deterministic code due to counting limitations and cost, while briefing generation (creative segmentation and narrative writing) was a natural fit for LLMs. The agentic approach used web search and fingerprint summary tools to enable multi-turn responses instead of a single overloaded prompt.

Parker Conrad knows which employees are worth their AI spend and says Rippling can help you, too

Parker Conrad demos Rippling Data Cloud, a new product that consolidates HR, business intelligence, and cross-system analytics into one platform. A key use case shown is tracking AI token spend per employee and correlating it with performance metrics — Rippling found some employees spending $30K/year on Claude with little ROI, and engineers with high AI spend but high code review rejection rates. The platform pulls data from sources like Anthropic usage logs, GitHub PRs, and Salesforce to surface workforce insights without a traditional data stack. Rippling also launched Business Banking with same-day payroll. Conrad says the company is ~2 years from cash-flow positive and has no IPO plans.

​"พิมพ์เขียวเชิงโครงสร้าง (Enforcement Blueprint) ใน Step 9: Utilization…

A walkthrough of applying the Big Data Analytics Lifecycle (9 steps) to the Northwind dataset to detect and fix discount leakage in sales data. The analysis reveals a 15.9% discount leakage rate and identifies specific salespeople with anomalously high discount rates. A Scikit-Learn ML pipeline simulates discount cap policies (5% and 8%), showing potential margin recovery of up to $1,817.94. The final step (Utilization) implements an automated real-time transaction enforcement system that rejects discount violations, framing data governance as a structural discipline for protecting net margin.