Urgent.News

What's breaking now, across thousands of outlets.

AI

Mind Discipline: Why Our AI Advisor Only Reads Hand-Crafted Contracts

In my first post, I wrote about why I spent my first week writing zero business logic and instead built rig - our lightweight, POSIX-compliant local provisioning tool. It was my way of rejecting "wiki-ops" and applying Infrastructure-as-Code (IaC) discipline to our local environments so that a hardware failure means minutes of downtime, not a week. But as I transitioned into Week Two, I was hit…

In his second account, the author recounts a pivotal moment when their Confluence documentation was unexpectedly deleted, serving as a wake-up call about the importance of keeping documentation alongside code. This realization led to the adoption of a new architectural philosophy: treating documentation as a contract, stored in a unified, git-backed format.

By doing so, the startup eliminated the need for a bloated, detached corporate wiki and created a zero-overhead, anti-AI AI strategy. This strategy utilizes GitLab CI/CD and Google Workspace to build a secure, managed Retrieval-Augmented Generation (RAG) pipeline. The central idea behind this approach is to prioritize human intent over AI generation, treating AI as a principal architect and advisor rather than a code generator.

To achieve this, the team drafts several markdown files within the project repository: README.md, INTERFACE.md, and API.md. These documents serve as the source of truth, defining the component's internal mechanisms, public-facing boundaries, and exposed ReST endpoints. The author emphasizes that if a public endpoint or module contract isn't documented in the API.md file or the interface contracts aren't exposed in INTERFACE.md, then, to the rest of the system and the AI advisor, it does not exist.

This approach ensures deep clarity of thought before code execution and promotes a more intentional, hand-crafted software development process.

Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at dev.to →

More in AI

How to Build an AI Employee With a Knowledge Graph (Not Just Another Agent)

An AI agent can take an action. An AI employee needs to know what happens next. Most AI agents look something like this: Think → Act → Observe → Repeat That's fine for short-lived tasks.

  • Model the AI employee's tasks with nodes and relationships in a knowledge graph.
  • Introduce events and scheduling to make the AI employee functional and aware of task timing.

Your Free AI Server Will Fail Quietly. Five Gates to Make It Loud.

Your Free AI Server Will Fail Quietly. Five Gates to Make It Loud. The model can be innocent. The server cannot. Earlier this week I wrote a fail-closed checklist for AI-generated code.

  • Free AI server setup mirrors desired infrastructure but fails silently.
  • Five gates implemented to make failures more noticeable and detectable.
  • Gates include kill switch, budget counter, timeout, output validation, and health probe.

More from Friday 28 August →