The Future of Coding Agents
This is a victory-lap essay that turns into a forecast. It opens mid-celebration — "It has been three days since I launched Gas Town!" — and then Yegge steps back to ask where the whole field is going once agents stop being solo assistants and start being workers in a swarm.
The thesis: nature prefers colonies
Yegge's core move is to stop optimizing the individual agent. When there's real work to do, he argues, the winning unit isn't a genius loner — it's the colony. He relays a friend's line that Claude Code is "the world's biggest fuckin' ant": astonishingly capable, but built to run in numbers under coordination. The consequence for tooling is specific — agents will need to be born knowing how to be orchestrated.
"coding agents are very soon going to shift focus to be better colony workers. They need to have direct, built-in support for the emerging 'Orchestrator API Surface'"— The Future of Coding Agents
Free upgrades: why agents get better whether you help or not
Under the heading "Gas Town's Free Upgrades," Yegge lays out why coding agents improve almost automatically in 2026: the base models keep getting stronger, agent behavior gets baked into training data, vendors compete to make their agents better factory workers, and the community piles on contributions. Gas Town itself, he notes, was his fourth orchestrator — the design converged through iteration, not a lucky first draft. And he wrote it in Go on purpose: simple, legible, and easy for agents to keep straight, which is the actual point of "Why Golang?"
Factories win
The forecast has a hard edge. Colonies, factories, automation — Yegge doesn't hedge on which way this goes, and he's blunt that the economics are not up for debate.
"Factories are going to win. Automation is going to win. Of course they're gonna fucking win."— The Future of Coding Agents
His name for the state of play is deliberately provocative: this isn't artisanal craft anymore, it's industrial production — and, he insists, it's a blast.
"This is it. There's no going back. We've arrived at factory farming code. And it's hella fun."— The Future of Coding Agents
The future of big companies
Under "The Future of Big Companies," Yegge argues orchestration doesn't just change how you code — it destabilizes the org chart. A small team wielding a colony can out-produce a giant, so the giants have too many people doing too little work. He predicts a churn before things settle, and a new norm: in a world where a few operators run swarms, everything has to be visible or it's obsolete.
"The entire world is going to explode into tiny companies, which will then aggregate and re-form into larger ones… but not until we go through at least a year of churn"— The Future of Coding Agents
The gate: you have to be good enough to see it
The essay ends on a catch. Orchestration isn't an entry-level tool — it only pays off once you already have the skill to drive it, which is why most people who glance at Gas Town won't get it yet. Yegge puts a number on the threshold, borrowing an XP-and-levels frame from earlier in his writing.
"You need to be at least level 6, and have about half of the XP needed to reach level 7, before you'll appreciate Gas Town."— The Future of Coding Agents