Each quarter the firm reads AI labor demand from a proprietary database of roughly 700,000 job postings, patented and in production since 2018, and publishes where the return is landing and what the companies collecting it are building.
The first edition publishes with its methodology attached, reporting what the companies collecting a return from AI are hiring for and building, what the rest are still advertising for instead, and what the difference implies for capital already committed. Every figure carries its source and its denominator. Alongside it: an argument for why one weak factor can hold an entire program near zero, and an examination of the last time a general-purpose technology forced companies to rebuild the work around it.