Original
Long-form technical analysis written for engineers and architects who want substance over summary. Topics span model architectures, production deployment patterns, retrieval systems, testing methodology, and AI infrastructure.
Featured Study
An independent vibe coding benchmark comparing Claude Code, Cursor, and OpenAI Codex on a single prompt to create a complete browser arcade game.
All Articles
A practical benchmark comparing Claude Code, Cursor, and Codex after asking each to generate a complete single-file browser arcade game.
Weighted scoring, output analysis, and practical recommendations after comparing four AI-generated timeline diagrams created from the exact same prompt.
Methodology, scorecard results, and remediation ideas from an independent benchmarking effort designed to test whether major LLMs treat US framing as global truth.