My Foundation Shook a Bit

I’ve been using AI hardcore since just after the ChatGPT chatbot hit the interwebs in November of 2022.

Since then, I’ve tried many of the models and tools, and now use a combo in my daily workflow. I’ve moved on from the simple chatbot interface (which I still use quite a bit, however) to agentic coding agents and now build my own agents with agents.

These pages attest that I’m at least an AI apologist, and I might even categorize myself as an accelerationist. I’ve not been worried about AI taking over. I’ve not been worried about nefarious AI. I’ve not been worried about the Player Piano scenario, whereby AI does all the work and we sit around doing nothing.

Well, my foundation shook a bit this week.

I built a “systems engineering agent” (based on Claude Opus 4.6). I told it how to do its job from on my experiences, best practices, and industry. About 45 minutes later, “he” was ready. I fed it our marketing requirements document and told it to spit out a systems requirements document for a particular feature set.

Less than an hour later, I had a document that was 80% there. That was at least 2 weeks of work for an experienced human.

Two weeks => less than an hour.

Never more than right now do you need to understand that either AI will for you, or you will work for AI.