I write once I have figured it out.

What worked, what did not and what it cost. AI without the magic, code and infrastructure breakdowns, experiments with numbers.