Building with LLMs and AI-assisted development in real codebases.
View All Tags
An automated AI review passed front-end work that had missing interactions, no hover states, colors that did not match the design and a laggy page. I think the LLM reviewer is wrong in its concept. It should work like a QA engineer and check the implemented page against the design, so I set up a front-end testing procedure that validates the work from the mockup, step by step.
Vibe coding an AI agent means you tweak a prompt, rerun it and ship when the output feels right. The feeling is the only check, and it does not run in CI. I build agents the way I build any other software: modular code, mocked dependencies and tests written before the logic.
Asking a Drupal site which PHP version it runs should not require an SSH key and a terminal. I built Project Context Connector, a Drupal module that exposes a read-only JSON snapshot of site configuration to AI agents, automation scripts and monitoring tools. It offers four interfaces: REST, an HMAC-signed endpoint, Drush and native Model Context Protocol support.
AI made finding vulnerabilities cheaper. It did not make fixing them cheaper. The cost lands on an open-source maintainer with limited hours, who now has to sort a flood of security reports into urgent, false positive and later. Drupal AI Vulnerability Guardian works on that sorting step: it scans Drupal modules for 12 vulnerability patterns and gives each finding a score, a CWE ID and an estimate of the work to fix it.
Version one of Drupal AI Content Impact Analyzer returned one opaque number, which told editors how a page scored and nothing about why. It was a minimal proof of concept with a single scoring function, no tests and no documentation worth mentioning. It proved the idea worked but not that editors could use it, so the upgrade rewrites the core into a 6-dimension scoring engine that tells them what to fix.
The first version of this Drupal agent could create, read and list nodes. That proved the concept, and it could not fix a typo in the content it created. A content operations agent also needs update, delete, users, taxonomy and media, and it has to survive a flaky network, so retry logic now lives once in the HTTP client.