Engineering stack
Engineering stack
Featured content
Blog
Black box reverse engineering
Usually, one of our approaches to accelerating legacy modernization with AI is to use it to accelerate the reverse engineering part first, feed AI with the existing code and then let it help us create a comprehensive description of the application’s functionality, which can be used in the forward engineering. But what if we don’t have the code, or it’s so messy that it’s useless? How can Generative AI accelerate reverse engineering in this case?
Blog
The five dimensions of AI adoption
Adopting AI successfully requires more than technology; it demands a holistic approach built on five dimensions. It starts with developing AI literacy, from prompt engineering skills to critical thinking and architectural awareness. Engineering excellence, underpinned by practices like test-driven development and automation, ensures reliability and mitigates bias. Communities and champions help embed AI across teams, while governance provides guardrails without constraining innovation. Experimentation, grounded in business-driven hypotheses, prevents fragmented or short-lived efforts.
Blog
AI evals for MCP in AIOps
As AI agents in operations shift to autonomous decision-makers through the Model Context Protocol, they gain powerful access to logs, metrics, traces, and runbooks, along with new risks. AI evaluations embedded in these workflows provide essential trust, safety, and accountability. They combine automated metrics, LLM-driven reasoning checks, and expert human reviews to ensure relevance and reliability. This layered approach delivers visibility and governance for complex operational decisions. The result is AI that supports operations with confidence, safety, and oversight.
Blog
How to evaluate an LLM system
Evaluating an LLM system begins with crafting a robust, human curated ground truth dataset that reflects real user queries and domain context. Organizations then define precise evaluation metrics such as relevance, coherence, and contextual accuracy tailored to specific use cases. Comparing model outputs against this ground truth enables metric driven adjustments across prioritization, chunking, retrieval, and prompting strategies. Using stronger LLMs as evaluators can improve efficiency, but human evaluations remain the most reliable standard.
Hands on engineering knowledge
Detailed descriptions of experiments at the cutting edge of AI first software delivery
The faces at the forefront of technology innovation
The faces at the forefront of technology innovation
What’s new on martinfowler.com
Martin Fowler, Chief Scientist at Thoughtworks, is a world renowned author and speaker, and a pioneer of agile software development.
Article
How far can we push AI autonomy in code generation?