When I was at AWS, we measured everything. Page response times, API call patterns, user session durations. The metrics were beautiful in the way that spreadsheets are beautiful — clean, orderly, reassuring. What we never measured was what we were taking.
Every API call has two sides. One side is the service delivered. The other side is the data captured. In the early years of cloud, the focus was on the service. Over time, quietly, the incentive shifted toward the capture.
This is not a conspiracy. It is a structural outcome. When a company's revenue model is tied to advertising, and advertising revenue is tied to behavioural prediction, and behavioural prediction is tied to data, then every engineering decision gets filtered through that lens. You don't need to be evil. You just need to follow the incentives.
What "optimisation" actually means
In engineering culture, optimisation is a term of pure praise. To optimise something is to make it better. But better for whom?
When social platforms optimised their notification systems, they weren't optimising for user wellbeing. They were optimising for click-through rate. When recommendation algorithms were tuned, they weren't tuned for accuracy or satisfaction — they were tuned for time-on-site. These are not the same things, and the gap between them has been quietly colonised by engagement mechanics that most people never consented to and never knew existed.
I worked on systems that felt, from the inside, like genuine improvements. Faster load times. Better uptime. Lower latency. These were real. But they were in service of a machine whose purpose, at the highest level, was to turn human behaviour into a product.
The infrastructure choice is a values choice
When the Government of Canada and Telus announce a billion-dollar AI data centre cluster and call it "sovereign AI infrastructure," the word sovereign is doing a lot of work. Sovereign for whom? The data flowing through those facilities will be processed by systems designed by engineers working inside incentive structures I recognise. The sovereignty being described is national — it means Canadian data stays in Canada. That is not the same as saying Canadian people benefit from it.
The architecture of extraction doesn't change because it crosses a border. It changes when the incentive structure changes.
What would different look like?
Different would look like infrastructure built with public governance. Where the questions — who benefits, who pays, who decides — are answered before the concrete is poured, not after. Where the communities living next to the data centres are at the table, not reading about the announcement in a press release.
Different would look like slower, more intentional technology. Built to serve specific human needs rather than to maximise surface area for data collection.
That's what the Idea Bank is for. Not to protest the current architecture, but to make the alternative visible.