AI Blog

Deep Technical Writing

Author	Field	Characteristics
Hamel Husain	Agent evaluation, error analysis, and AI product improvement	Starts with production traces and human review, builds failure taxonomies, and turns high-value checks into automated evaluations.
Eugene Yan	AI product engineering, evaluation systems, and reliability	Distills practical experience into reusable methods, with a focus on eval-driven development, engineering workflows, and compounding knowledge.
Vicki Boykis	ML systems, agent-assisted programming, and human-AI collaboration	Examines generated code from a working engineer's perspective, emphasizing system understanding, cognitive load, software craftsmanship, and long-term maintainability.
Sebastian Raschka	LLM architecture, coding agents, and local open-source tooling	Combines clear technical explanations with working implementations. His agent-related writing focuses on coding harnesses, local open-weight models, and tool architecture.
Nathan Lambert	Open models, post-training, agent capabilities, and the AI ecosystem	Connects technical and industry analysis, with a strong ability to assess model capability boundaries, practical agent thresholds, and shifts in open and closed ecosystems.
Vincent Warmerdam	Python tooling, notebooks, data, and experimental workflows	Writes concise, experiment-driven, reproducible posts. Consistently emphasizes understanding the data and problem before selecting models or automation tools.
George Hotz	AI automation, multi-agent systems, low-level systems, and technology economics	Connects engineering, hardware, business, and social impact. His views are sharp and original, but best read critically.
Ben Recht	Machine learning evaluation, optimization, control theory, and AI criticism	Questions mismatches between benchmarks, objective functions, and product value, helping readers identify hidden assumptions in AI systems.
Lilian Weng	Agent architecture, reasoning, self-improvement, and AI safety	Publishes infrequently, but each article is technically dense and typically synthesizes the literature on agents, harness engineering, and capability development.

Technical Radar

Author	Field	Characteristics
Simon Willison	LLMs, agent engineering, tool use, security, and open-source ecosystem	A high-frequency technical intelligence source: tracks primary announcements, papers, code, and industry events, then adds experienced engineering context and useful skepticism. Best for discovery and judgment rather than as a steady source of deep, evergreen technical articles.

Appendix: Supplementary Reading

Author	Field	Why it is supplementary
Alex Lavaee	Coding-agent workflows, context engineering, and agent infrastructure	Frequent, practical writing with useful workflow templates and implementation ideas. Read selectively: engineering advice is more reliable than its broad theory, model-comparison, or trend claims; verify primary sources before relying on those claims.