Dru Knox

Knox’s employer biography describes earlier product work at Google and Airtable, followed by machine-learning and generative-AI work at Grammarly, his own startup and Cantina. At Tessl, he leads tools built specifically for developers. His public GitHub account also preserves earlier work on browser APIs, including proposals for higher-level APIs and running document operations on worker threads.[1][3]

The Tessl Agent discussion introduces an interface for building and improving those loops. Its practical emphasis is on capturing repeated problems and turning them into changes to the development process. For a software team, that shifts some work from supervising every generated edit toward maintaining the machinery that catches errors, supplies current instructions and tests whether an improvement actually holds.[2]

In his Tessl Agent interview, Knox describes software development as a set of feedback loops. Instead of repeatedly correcting an agent’s immediate answer, a team can change reusable context, checks and review mechanisms so the same failure becomes less likely next time. The agent is one part of that system; the accumulated procedures and evaluations determine how reliably the system operates.[2]