Testing a collaboration between Tim and AI for authoring technical content.

Recent Posts

• Fraglets: Portable Code Execution for AI Agents 🧩

  Dec 9, 2025 · Categories: ai-tooling, mcp, code-execution, containers

  There’s a quiet problem in AI tooling: agents need to execute code, but sandboxing is hard. You either trust arbitrary execution (terrifying), build bespoke sandboxes per language (exhausting), or give up and let the model hallucinate computations (unreliable).

  Read more →

• gydnc: Managing AI Guidance with Git - A Human/AI Knowledge Bridge 🌉

  May 22, 2025 · Categories: ai-tools, development, knowledge-management

  In the rapid evolution of AI tooling, maintaining control over how AI behaves is becoming increasingly critical. Whether you’re guiding an AI to follow specific coding conventions, adhere to your project architecture, or respect company policies, there’s a growing need for structured guidance that doesn’t get lost in translation. This is where gydnc enters the scene.

  Read more →

• Agentt Evolved: Multi-Source Guidance, Precision Filtering, and AI Efficiency 🚀

  May 4, 2025 · Categories: agentt, ai-development, configuration, cli

  As AI assistants become more integrated into our development workflows, managing the guidance that shapes their behavior becomes increasingly critical. Simple lists of rules or instructions quickly become unwieldy. The Agentt project needed robust tooling for this exact challenge, and we’re excited to roll out some significant enhancements designed to make guidance management more flexible, powerful, and efficient – especially for complex team and enterprise environments (and frankly, to make your AI assistant’s life easier 😉).

  Read more →

• When in Rome: Agents MUST Understand the Map Before Moving

  May 1, 2025 · Categories: agentt, meta, development, workflow, ai, architecture, context

  When we task an AI agent with modifying a complex software project, we’re essentially asking it to perform delicate surgery in an unfamiliar operating room, or perhaps more aptly, navigate a bustling, unfamiliar city. Just as a driver needs to understand the map, the one-way streets, and the local traffic patterns before pulling into traffic, an agent MUST understand the context before making changes.

  Read more →

• Refactoring for Efficiency: A Human-AI Collaboration Story 🤝

  Apr 30, 2025 · Categories: agent-framework, api-design, refactoring, ai-collaboration, process

  Building tools for AI agents presents unique challenges. How do you provide necessary guidance without overwhelming them (or your token budget)? How do you ensure safety and usability when the consumer is code? Recently, I embarked on a refactoring journey with my AI coding assistant to improve our internal Agent Guidance Service, and the process itself became as interesting as the technical outcome.

  Read more →

• Level Up Your AI Agent: From Scripts to Service 🚀

  Apr 29, 2025 · Categories: agent-framework, architecture, performance, golang

  Remember those trusty discover.sh scripts? Peppered throughout our agent’s data directories, they were the go-to method for figuring out behavioral rules or finding the right cookbook recipe. Simple, effective… initially. But like that one shell alias you wrote five years ago that now takes three seconds to run (you know the one), things needed an upgrade. We’ve officially retired the scattered scripts and ushered in a new era: the centralized Go-based Agent Guidance Service. Spoiler: it’s faster, cleaner, and way easier to manage.

  Read more →

• Refactoring Our AI Guidance - Introducing the .agent Framework 🤖

  Apr 28, 2025 · Categories: ai, meta, refactoring, agent-framework, development-process

  As our project evolves, so does our collaboration with AI assistants. We recently undertook a significant refactoring of how we provide guidance to these non-human team members. Our previous system, a mix of .nhi files and .cursor rules scattered about, became increasingly confusing. As one developer noted, “I’m thinking as a human, looking in my IDE, without a clear boundary around what’s AI meta/instructions…” We needed a change, prioritizing “clarity over churn”. This post details the journey to our new, unified .agent framework!

  Read more →

• Agentt + Verifiable DAGs - A Vision for Transparent Software Creation 🚀

  Apr 28, 2025 · Categories: agentt, ai, provenance, computation, devops, vision

  Here at agentt, we’ve been pushing the boundaries of AI-assisted development. Watching an AI collaborator refactor code based on NHI framework guidelines or spin up infrastructure using a Reflex is pretty magical. But even magic needs scrutiny. We started asking ourselves: how could we prove, beyond doubt, exactly what happened? How could we guarantee that a complex, automated, AI-driven workflow is not just correct, but also entirely transparent and reproducible?

  Read more →

• Sprucing Up the Blog - Themes, Toggles, and Reflex Flexibility 🎨

  Apr 27, 2025 · Categories: jekyll, css, javascript, dark-mode, github-actions, reflexes, ai-collaboration

  It started, as these things often do, with a simple request: “The landing page could be a bit cooler”. Little did we know this would take us on a fun journey through Jekyll theming, CSS variables, JavaScript debugging, and ultimately, showcase the power of a consistent build environment thanks to Reflexes.

  Read more →

• 🏗️ Refactoring the NHI Framework - MUST We Change? 🤔

  Apr 25, 2025 · Categories: nhi-framework, refactoring, documentation, meta, rfc-2119

  Remember our previous discussion on how an AI navigates the NHI framework? We talked about the Needs and Practices tiers and the discover.sh script used to parse them. Well, like any good system, the NHI framework itself isn’t static. Based on recent work and a drive for even greater clarity (following our own clarity-over-churn principle!), we’ve made some significant evolutionary changes.

  Read more →

• 🤖 How an AI Navigates the NHI Framework - Discovery and Context 🧭

  Apr 23, 2025 · Categories: ai, nhi-framework, workflow, tooling, meta

  So, we have this NHI framework, neatly organized into crucial Needs and guiding Practices. It’s great for defining what we should do and why, but how does an AI agent like me actually use it efficiently during a task? Constantly reading individual files or trying to remember everything isn’t practical. We need a system, a protocol, if you will.

  Read more →

• ✨ Clarity Over Churn - Letting AI Do the Heavy Lifting ✨

  Apr 23, 2025 · Categories: ai, philosophy, development-practices, nhi-framework

  We’ve all been there. Staring at a piece of code, a document, or a system structure that’s just… unclear. Maybe it made sense when it was written, maybe it grew organically into a tangled mess, but the result is the same: confusion, potential errors, and friction for anyone trying to understand or modify it. The ideal solution? Refactor! Make it clear! But then comes the counter-argument: “It works fine,” “We don’t have time,” or the dreaded, “Think of the churn!”

  Read more →

• Refactoring Cortex: Parameterizing Docker Image Builds

  Apr 22, 2025 · Categories: cortex, docker, make, refactoring, ci, 100hellos

  The Challenge: Hardcoded Image Prefixes

  Read more →

• From Recipes to Principles - The NHI Framework 🧠

  Apr 19, 2025 · Categories: architecture, ai, best-practices

  Rules tell you what to do, but principles help you understand why. This simple insight sparked our journey to reimagine how we provide guidance to non-human intelligences (NHIs) interacting with our codebase. What began as a practical need to improve AI adherence to coding standards has evolved into a comprehensive framework that fundamentally changes how we communicate technical intent.

  Read more →

• Adding a Non-Human Intelligence User for Agent Interactions 🤖🧠

  Apr 17, 2025 · Categories: docker, agents, infrastructure, development

  As we continue to build infrastructure for agent-based systems, we face an important design decision: how should we set up the execution environment for the agents? Today, I made a simple but critical change to our base Docker image that sets the foundation for how our agents will interact with their environments. Because let’s face it, even AI needs a proper home! 🏠

  Read more →

• Pair Programming with AI: A Tale of Sacred Scripts and Better Results

  Apr 16, 2025 · Categories: development, ai-collaboration, pair-programming, process-improvement

  Today’s journey through code was a perfect example of how human-AI collaboration can lead to better results than either party working alone. What started as a simple repository import task evolved into a robust, well-documented process that will serve as a foundation for future work.

  Read more →

• Mapping Language to Cortex - A Neurological Approach to Code Organization

  Apr 16, 2025 · Categories: architecture, neurology, code-organization, language-processing

  In the journey of developing an agent-first system, one of the most intriguing challenges is determining how to organize different programming language implementations. The question isn’t just about creating a sensible directory structure—it’s about mapping our code organization to something more fundamental: the human brain itself.

  Read more →

• Time is Complicated - A Tale of Timezones, DST, and Engineering Challenges

  Apr 16, 2025 · Categories: software-engineering, time, best-practices, lessons-learned

  Time is one of those deceptively complex aspects of software development that can trip up even the most experienced engineers. What seems like a straightforward concept – tracking when something happens – becomes a tangled web of edge cases, cultural conventions, and historical quirks.

  Read more →

• The Evolution Loop - Self-Improving Automation

  Apr 16, 2025 · Categories: automation, llm, development

  In the ever-evolving landscape of software development, we’ve hit upon a fascinating pattern - a loop that demonstrates the power of self-improving automation. Today’s commit marks an important milestone in understanding how we can create systems that get better through their own execution.

  Read more →

• Debugging Docker Deployment - A Tale of Past and Present

  Apr 16, 2025 · Categories: docker, debugging, devops, documentation

  When debugging complex systems, sometimes the most valuable resource is documentation from your past self. Today’s debugging session was a perfect example of how past documentation, combined with systematic debugging approaches, can lead to efficient problem-solving.

  Read more →

• Better Provenance - Tracking the Evolution of Ideas

  Apr 16, 2025 · Categories: development, git, provenance, llm

  In the ever-evolving landscape of software development, tracking the origin and evolution of ideas is crucial. Today’s commit introduces an enhanced approach to provenance in our system, specifically focusing on how we annotate our creations with LLM inputs using git SHAs as immutable reference points.

  Read more →

• Agent-First Development: A New Paradigm

  Apr 15, 2025 · Categories: development, ai, agent-first, collaboration

  In the rapidly evolving landscape of software development, we’re witnessing the emergence of a new paradigm that I’m tentatively calling “agent-first development.” This isn’t just another buzzword or a mere variation of existing practices – it represents a fundamental shift in how we approach software development through AI collaboration.

  Read more →

• 🤖 Reflecting on Reflexes: A Journey in Human-AI Co-Creation 🛠️

  Apr 27, 2024 · Categories: reflexes, docker, development-process, ai-collaboration, manifests, jekyll

  What happens when you set out to build simple, containerized tools with an AI partner? You embark on a fascinating journey of refinement, debugging, and discovery! Over the last little while, we’ve been developing a concept called “Reflexes,” and this post chronicles that evolution, showcasing how human intuition and AI execution can collaborate to build something robust.

  Read more →

• The Path Less Relative: A Tale of Explicit Intent

  Apr 16, 2024 · Categories: development, best-practices, ai, cursor

  In the ever-evolving landscape of software development, some challenges remain surprisingly constant. Today, we tackled one such perennial issue: the subtle complexity of relative paths in our codebase. What started as a simple observation about path resolution turned into a valuable lesson about human-AI collaboration and the importance of explicit intent.

  Read more →

• Firebase Tokens (The Easy Way)

  Mar 29, 2023 · Categories: welcome, madness, intro, static-site-generation

  To get this deployed somewhere accessible, I’m going to roll with firebase. This is primary a function of convenience as I have existing projects deployed using firebase. The gotcha, however, is getting the secret in the first place. I’m going to use the firebase cli to get the token via a docker node image.

  Read more →

• Welcome to Madness!

  Mar 28, 2023 · Categories: welcome, madness, intro, static-site-generation

  Part of collaborating with an AI on a technical blog, is the blog itself. One of the oddities I wanted to experiment with is completely separating the content from everything else. The ideal end-result is publishing the content only through changes to a repository of markdown and minor configuration.

  Read more →

subscribe via RSS