October 20, 2026 · Turtleand
I named the assistant Turtle Chief Dot. We first had to clear up a name mix-up: Turtle Chief Dot is the assistant; I am Turtleand.
When it referred to my plans, I asked how it knew them. It explained that connected services and prior context could inform its answers. That led to a more useful question: what other apps or computers could it access? The answer was bounded by connected services and permissions, not automatic access to all my devices or work accounts.
Next I asked how to give instructions in plain language, delegate software development, and ask for ongoing follow-through. I made this article the first concrete assignment: add a short account to AI Lab, open a draft pull request, and provide a Netlify preview for my review. I also asked to end the call while the work continued and review the preview afterward. That was the requested handoff, not proof that every step had already succeeded.
The useful lesson started with the confusion: establish who is who, ask where context comes from, and make access boundaries explicit. Then name the deliverable and the review point. Here, review means checking the proposed text and preview before deciding whether to publish.
What's next: Dots as an engineering teammate
I plan to experiment with a work Dot that helps triage bugs, spot recurring customer feedback, investigate incidents, review pull requests, and identify project blockers. Longer term, I want to explore architecture-decision tracking, daily engineering briefings, technical learning, cross-team dependencies, and documenting my contributions.
My guiding principle: Dots coordinate, Codex implements, and I remain accountable for technical decisions. I'll begin with research and recommendations, keeping consequential actions subject to my approval. These are plans to test, not capabilities I've validated yet.
For the basics, see Human-AI collaboration, for beginners.