After AI Takes Over Engineering, Measurement Gets Priced First

On July 28, Weave announced a $13.5 million Series A led by Standard Capital, with Y Combinator, Moonfire, Burst Capital, IrregEx, and the Agent Fund following. It came a little over a year after its $4.2 million seed round. Weave uses LLMs and its own models to analyze every pull request and quantify how much work engineers and AI each completed. When announcing the round, founder Adam Cohen said that AI has taken over engineering, that companies are wasting millions of dollars, and that Weave lets you get the most out of every dollar of token spending. This is not the first time the category has been priced. In September 2025 Atlassian announced it would acquire DX for $1 billion, a developer productivity measurement company with 350 enterprise customers, nearly all of them already Atlassian customers. In ten months the same category went through a $1 billion acquisition and a new Series A. The category's name isn't settled yet. Some call it engineering metrics and some call it engineering intelligence. Measurement has always existed in engineering management, but it was closer to a management preference, with tools optional and each team using its own methodology. Now it is being priced on its own. What changed is the thing being measured, and the measurement technology is the same. The demand for measurement comes from a failure of perception Spending on AI coding tools is no longer small. Cursor, Copilot, and Claude Code have become defaults on engineering teams, with purchase first and an answer about effect later. Up to now that answer has mostly been given by feel. The gauge of feel has a problem of its own. In the randomized controlled trial METR published in July 2025, 16 experienced open-source developers worked on 246 real tasks. The group allowed to use AI was 19% slower, and afterward they estimated they had been 20% faster. The study is small, and its conclusion may not hold everywhere, but the problem it points to is structural: the people using these tools can't perceive their effect, and the direction of their perception may be backwards. There is a similar signal on quality. GitClear's annual code analysis shows that after AI became widespread, the amount of code cloning rose to about four times what it was. When a spending line grows and perception can't calibrate its effect, a measurement layer usually grows around it. Advertising spend grew attribution and monitoring, cloud spend grew observability, and token spend has arrived at the same spot. Weave coined a name for it, tokenmaxxing, meaning the behavior of producing more code by piling on tokens without an equal contribution to the product. In Adam Cohen's words, the era of tokenmaxxing is over. The lead investor adds weight to the judgment. Standard Capital was founded by Dalton Caldwell, Paul Buchheit, and others, with a first fund of $425 million. It positions itself as an AI-native Series A investor, backing about 5 companies per cycle and requiring at least 10% ownership. Caldwell's account of this round was that AI spend is the strongest force in the world right now and there is no convenient way yet to measure it. Code now has two kinds of authors, and the last generation's data model breaks