For 15 years, Clojure's headwinds (smaller hiring pool, thinner library long tail, the inertia of moving off an incumbent stack) outweighed its ergonomic benefits (REPL-driven iteration, immutable data, image introspection, succinct DSLs) for most teams — the right call while humans typed. Agentic coding changes both sides of the scale at once: the AI absorbs most of the headwinds (the surface tax is paid by something that doesn't tire), while Clojure's ergonomics multiply a developer that iterates orders of magnitude faster than a human can. After starting a production AI product on TypeScript and Python (~90,000 lines) and rewriting it in ~12,000 lines of Clojure and ClojureScript with an AI agent as the primary pair, the rewrite paid for itself in months and has compounded gains daily ever since. You'll leave with a framework for re-running that calculation in your own setup — and a usable answer: don't let the agent dictate your language. The agent can handle yours.

Brendan Foote

Brendan Foote co-founded JoinIn.ai, where a two-engineer team rewrote a production real-time AI meeting product from 90,000 lines of TypeScript and Python into 12,000 lines of Clojure and ClojureScript — the case study of this talk. He has been working in Clojure since 2013, when he read The Joy of Clojure and gave a "Why Clojure?" presentation in his Master's in Software Engineering program. He has spoken on Clojure at MSSE Barcamp events, including a talk on a production Clojure tool from a prior financial-services role — 87 lines of code that parsed 750MB of nightly XML and trivially parallelized the work. With 20+ years of engineering experience, he has been making the case for Clojure for over a decade, and is making it again — now as an applied lesson from rebuilding a production AI product with an agent as the primary pair.