The Organizational Immune System Just Got Switched Off

Harrison Chase, creator of LangChain, posted this week on how coding agents are reshaping engineering, product, and design. It's worth reading. It's also built on a premise that quietly does too much work.

The one thing it genuinely gets right: bad product thinking is now actively dangerous in a way it wasn't before.

Chase puts it in a single paragraph and moves on, but it's the most important point in the piece. A poorly-conceived idea used to die as a whiteboard sketch — killed by the cost and friction of actually building it. That friction was an organizational immune system. Now the same idea shows up on Monday as a functional prototype. "It already exists — let's just merge it" is a real pressure, and it compounds. He names it and doesn't develop it. Someone should.

The thing the piece doesn't earn: "anyone can write code now."

That claim carries the entire argument. It's the reason the bottleneck shifts, the reason generalists win, the reason PRDs are dead. But Chase never stress-tests it. Anyone can generate some code for some problems with some probability it works. The variance is enormous. It's fine for a greenfield CRUD app at a seed-stage startup. It falls apart fast against a distributed payments system, a legacy codebase with fifteen years of load-bearing workarounds, or anything that needs to survive a security review. His "reviewer" archetype — the person who catches what agents get wrong — is just a senior engineer doing what senior engineers have always done. The bottleneck moved. It didn't disappear.

Chase's frame is accurate for the world he's in: AI-native startups, greenfield products, small teams. It generalises poorly. Most software that matters is running inside organisations where the delays were never caused by how long it took to write the code.

Read it for the bad-PM insight. Hold the rest loosely.