This is for people who only use chat AI, keep testing new AI tools, and worry about falling behind. You see a Zig headline, a Claude headline, a fight about AI code, and you almost scroll past. Then you stop because maybe this changes your next move. The useful part is not the drama. It is the bill hiding underneath it: the most expensive part of AI code is review minutes. That is the part that breaks the Claude myth.
Zig Creator Calls Spade a Spade, Anthropic Blows Smoke. Read that as a correction to the easy story that cheaper code automatically means cheaper software. If AI helps you on your own branch, it may save your own typing. Once that code turns into a public contribution, somebody else has to read it, test it, question it, and decide whether it belongs. That is why public open-source review is a different problem from private team use.
The Zig example matters because it shows where the bottleneck moved. Andrew Kelley said AI-assisted contributions could be "negative value" because they consumed core team review time, and Zig had about 200 open PRs waiting for review [S001]. That number matters more than any flashy demo. When writing code gets cheaper, the reviewer's judgment becomes the scarce resource. Faster generation does not remove the queue. It can make the queue worse.
A 2026 study across 294 open-source projects found the same pattern: more AI-made submissions, but fewer accepted changes [S005]. The paper describes this as AI-DDoS: code generation gets cheaper while evaluation gets more expensive. In plain English, one more patch is easy to produce and expensive to trust. That flips the real question from "Which model writes more code?" to "Who pays for reviewing all this code?"
Do not judge an update by how many features it lists. Judge it by whether it changes your next decision. Here, the next decision is who has to review the output, how long that review takes, and whether the queue grows faster than the value. Share this with the person who still thinks faster AI code automatically means lower software cost. Boundary: this argument is about public open-source review as of 2026, not private team use inside your own branch.