For twenty-six years, the full rewrite was the cardinal sin of software engineering. Then one engineer ported half a million lines of Bun from Zig to Rust in eleven days. The taboo was never about code. It was a price tag, and for systems whose intent lives in an executable oracle, the price just collapsed.
Satya Nadella named a real problem: AI workflows generate valuable learning exhaust from proprietary context and corrections. He is right about the asset. My practitioner's edit is simple: own the learning loop, not necessarily the model.
In June I itemized the AI subsidy by hand. This month I built the instrument that does it for me. It took an afternoon, and it was built by the very thing it measures.
Twenty-six days of Claude Code showed $2,556 of API-priced work against a $200 subscription. The lesson was not the total. It was cache behavior, model routing, and a government kill switch that landed in my usage chart. Value lives in verified outcomes, not tokens, and the work has to survive the stop.
Build, WWDC, two confidential S-1 filings, and a tiered frontier model launch hit in the same two weeks. The agent became the interface just as the invoice arrived.