Technical deep dives on AI development tools — Google ADK, OpenAI Agents SDK, Cursor, and Claude.
10 articles
I run nine AI agents in one live system on WhatsApp. Getting the topology right took an afternoon; keeping it alive is the job. A watcher that lost an instruction on restart, a watchdog that killed working agents at four minutes, silence that read as failure, an agent that approved its own work, and a credential read once at boot that could never be retried into success — all real, all mine.
Prompt a coding agent as a to-do list and it loses the thread, redoes finished work and drags early mistakes forward. I express the job as a dependency graph instead, and that is why a fixed price is possible. Plus the knowledge base that keeps each run from starting at zero.
"Our AI can't answer that" is usually a data-shape problem, not a model problem. What changes when your business is stored as entities and relationships instead of flat lists, why coding agents turned a six-month enterprise project into a few weeks, and the afternoon test that tells you whether you need one at all.
Nine times out of ten, "we want our own AI model trained on our data" describes a retrieval problem, not a training one. The four tests that decide it, what owning a model really costs after launch, and the five questions to put to anyone proposing to train you one.
One script becomes a multi-scene film on Google Veo 3.1, with scenes written by Gemini 3.7 Flash. The build, the constraints nobody warns you about, and what shipping bring-your-own-key taught me about pricing an AI product.
Agency OS is a white-label AI operations platform I built for agencies: your brand, your playbook, and your clients' data on your own server.
A practical comparison of the four major AI agent platforms in 2026. What each does and how to choose.
Cursor 2.4 introduces subagents, Agent Skills, background cloud agents and new enterprise features. What each one does.
A comprehensive guide to building autonomous AI agents using Google's ADK framework. Multi-agent orchestration and deployment.
Build production-ready AI agents that perform complex multi-step tasks using OpenAI's Agents SDK with handoffs and guardrails.
