We talk about AI as if it were only a capability. It is also a bill — paid in cash, in electricity, in water for cooling, and in the quieter currency of human attention.
Cost is not only the subscription line. It is the meeting that grew longer because the draft was easy to generate and harder to decide on. It is the junior who stops practicing judgment because the model answers first. It is the product that ships faster while the support queue fills with people who cannot find a human when something breaks.
Environment is not a footnote for someone else’s ESG report. Training and inference draw power and water at a scale most interfaces never show. That does not mean we refuse the tools. It means we stop pretending the cloud is weatherless. A healthy system names what it consumes.
Human is the part that cannot be automated without loss: taste, responsibility, the courage to say no, the willingness to sit with someone whose problem is not a prompt. AI can polish a sentence. It cannot bloom a life. At best it frees a little room for us to do the work that still requires a person.
I am not interested in being early for its own sake. I am interested in adoption that stays close to real life — where the cost is visible, the environmental load is not ignored, and the human remains the one who decides what “better” means.