After the feedback loop closes

Why AI takeoff can be fast without making the physical world instantaneous.

AI helping a researcher is one thing. AI improving the process that produces the next generation of AI is another. The takeoff scenario begins when that feedback becomes reliable enough to accelerate subsequent research materially.

The paper uses March 15, 2028 as a concrete starting point. This is a storytelling and analytical anchor, not a prediction that a laboratory crosses the threshold on that day. Readers can change the date while still examining the sequence.

Software and other testable digital work can respond quickly. The inherited physical world moves on different clocks: power connections, chip production, laboratory equipment, construction, and deployment all impose constraints.

But a bottleneck is not necessarily permanent. More capable systems can improve algorithms, experimental design, hardware design, logistics, and eventually the machinery used to expand capacity. The key question is how quickly intellectual improvements translate into physical throughput.

For households, the resulting changes need not arrive together. Asset values, wages, debt obligations, and the price of digital services can move in different directions. A family can become wealthier on paper while losing the income that supports its daily life.

The purpose of a conditional future history is to make those interactions concrete enough to dispute. It gives readers a sequence to test, a set of assumptions to replace, and a way to ask what would slow or accelerate transformation.