ideas

Ideas

I write about AI-assisted development, engineering leadership, systems thinking, and the messy human reality between them.

2025

  • 2025 Recap - The Year I Didn't Give a Damn About People

    Even though I never considered myself a great programmer, I always loved to code.

    4 min read recap-2025 · management · sabbatical · agents · claude-code · software-development · leadership
  • 2025 Recap - From Barely Walking to Training Three Times a Week

    Early in this bumpy Sabbatical Year, on top of being diagnosed with Sleep Apnea, I was diagnosed with Plantar Fasciitis, which I only recently managed to overcome after a year of physical therapy, shockwave therapy, dry needling, and more physical therapy.

    5 min read health · recap-2025 · plantar-fasciitis · training
  • 2025 Recap - My Life with Sleep Apnea

    At the start of 2025 I began a sabbatical after six years at full throttle in a startup.

    8 min read sleep · sleep-apnea · recap-2025 · cpap · polysomnography
  • Backpressure - Rejecting Negative Hallucinations in Programming Assistants

    In this fifth and final installment of the “Fundamental Principles for Using Programming Assistants” series we are going to make peace with — even embrace — the fact that no matter how many instructions and examples (steering) we hand an LLM, there is no way to keep it from generating code that is no good, that is: negative hallucinations.

    6 min read programming-assistants · negative-hallucinations · code · backpressure
  • All Prompting is Prompt Injection

    There is no such thing as “Prompt Injection” as an attack mechanism, nor as a vulnerability.

    1 min read prompting · prompt-injection · llms · tokens · security · attack-vector
  • Steering - Favoring Positive Hallucinations in Programming Assistants

    This fourth installment of the “Fundamental Principles for Using Programming Assistants” series focuses on “steering”, the term I will use here to refer to the practices that let a Programming Assistant consistently favor the production of positive hallucinations — hallucinations that deliver value to developers and to the users of our application.

    8 min read programming-assistants · context · code · vibe-coding · positive-hallucinations · systems
  • A Systems View of AI-Assisted Programming

    In the first article of this series we concluded that hallucinations are the fundamental feature of the LLMs on which Programming Assistants like Claude Code, Cursor, Gemini CLI, Windsurf, etc. are built, and that these hallucinations can be positive or negative depending on whether they add value to our code or not.

    6 min read llm · programming-assistants · context · prompting · systems · steering · backpressure · processes · flows
  • The Mathematics of AI-Assisted Code

    From our first article in the series, Hallucination is the Fundamental Feature of LLMs, we have that LLM hallucinations can be positive or negative depending on whether they add or subtract value from our work, from “our code”.

    3 min read llm · programming-assistants · mathematics · negative-hallucinations · positive-hallucinations · context · prompting · value
  • Hallucination is the Fundamental Feature of LLMs

    An LLM cannot not hallucinate because in essence that is how it works: hallucinating that next word (tokens), producing a plausible text (image, audio, etc.) that may have no grounding in reality.

    4 min read programming · positive-hallucinations · negative-hallucinations · hallucinations · software · llm · software-development · programming-assistants
  • Using a Github Page with your own Domain to publish you Obsidian vault through Quartz for free

    This is a detailed step-by-step tutorial that will allow you to set up Quartz to publish a folder of your vault as a website for free in Github Pages, is mostly a reminder for myself, I hope that it may be useful for you.

    11 min read github-pages · obsidian · quartz · macos · website · free