Juan G. Carmona

Topics

ai-sdlc

8 articles, most recent first.

  1. Define Once, Cite Everywhere

    Same rule. Three specs. Three wordings, and two different meanings. Thirty days from purchase is not thirty days from delivery. Nobody decided that fork. Nobody reviewed it. A coding agent paraphrased it into existence, because every new spec re-explains the product from scratch, and paraphrase is drift.

  2. I Tried to Prove PDaC Wrong

    I am shaping Product Definition as Code (PDaC) specification and building its first reference implementation, productshape, and this week I tried to prove it wrong.

  3. The Wiki Knows What We Learned. The Product Model States What We Mean.

    Someone recently told me that ProductShape is essentially the same idea as Andrej Karpathy's LLM Wiki. Is it? No.

  4. The Symbolic Layer Before the Spec: What Ontologies Teach the AI-SDLC

    A few days ago, I published Product Definition as Code for the AI-SDLC and ProductShape, its reference implementation.

  5. Product Definition as Code for the AI-SDLC

    Lately, I have been working as a FDE (Forward Development Engineer) at Plain Concepts, helping different clients introduce AI into their software development lifecycle. AI-SDLC and Spec-Driven Development are my two main areas of focus and I have been thinking a lot about the implications of AI-assisted development on the software lifecycle in many environments, with many tools, with different workflows in many different scenarios...

  6. What DevEx Means in 2026

    Over the last months I have been working with many organizations navigating the shift toward AI-assisted development. The conversations always converge on the same topics. Software engineers and leadership teams already understand that the software delivery model is changing, yet most organizations are still focused on the wrong question.

  7. The Missing Standard Behind AI-Assisted Development

    Over the last months I have been helping engineering organizations move their software delivery lifecycle toward AI-assisted development.

  8. Multi-Agent AI Repository Architecture: The Canonical Semantic Layer

    A Git repository used to serve humans, build systems, and CI/CD pipelines but today it also serves AI-tooling and Harnesses like Claude Code, GitHub Copilot, OpenAI Codex, OpenCode, Cursor, Aider, Gemini CLI, and whatever comes next.