The article you just read argues that working effectively with AI requires human-centric skills — philosophical thinking, adversarial instincts, linguistic precision, and creative resourcefulness. It felt only right to show how those same principles were applied in writing the article itself.
What follows is a transparent walkthrough of the multi-agent, multi-platform workflow used to develop "The Human Factor" from initial idea to published piece.
The tool: Google Gemini Flash (fast, lightweight model)
The task: Rapid-fire brainstorming. I came to this project with a loose collection of ideas and my own framework about what makes people effective in AI-driven work — observations from my own experience building multi-agent systems and automation workflows. At this stage, I needed speed over depth. I wanted to sketch the shape of the argument quickly, test different framings, and see which ideas had legs.
Why this model: Flash is built for speed. It's ideal for the messy, generative phase of thinking where you're producing a lot of rough material and discarding most of it. You don't need deep reasoning here — you need volume and velocity.
The tool: Google Gemini with extended thinking (deeper reasoning model)
The task: Two things happened in this stage. First, I worked with the model to develop the real-world examples that would anchor each section of the article — the logistics ecosystem, the security protocol, the ad campaign. I had to translate my own real-world but confidential business cases into something analogues but blinded. The challenge was translating specific, proprietary experiences into generalized examples that are honest and illustrative without exposing sensitive details.
Why this model: This required more sophisticated reasoning than ideation. I needed the model to help me think through which details were essential to the point and which were identifiable — a judgment call that benefits from a model capable of holding multiple considerations in tension. Gemini's extended thinking mode gave me the depth to work through that carefully. In general, I’ve found ideating with Gemini, often via voice chat, to be more frictionless (less frictionfull?) than the other models.
The tool: Anthropic's Claude Opus 4.6
The task: With the outline, examples, and masked frameworks in hand, I moved to Claude for the actual narrative drafting. This was the stage where the article went from structured notes to a readable piece with a consistent voice, pacing, and tone. I worked iteratively with Claude across multiple revision passes — adjusting tone (dialing back an initial draft that was too punchy), adding warmth and humor by having it match other of my works, and refining the examples until they landed naturally.