Codebase Grows Twentyfold — Programmers Trapped in AI Efficiency’s Cage

AI productivity — and the paradox of more work

Tech efficiency has a new face: AI that writes code. But faster delivery has not translated into less toil. It has been reported that Oracle (甲骨文), after reporting a 95% year‑on‑year jump in quarterly net profit, is both aggressively expanding AI datacenters and carrying out broad, performance‑agnostic layoffs. The result? Engineers are seeing output metrics balloon — code volume, pull requests, even AI token usage — and being judged against them. Who wins when the machine speeds up but humans are forced to keep pace?

Voices from the codebase

Programmers in the U.S. and China describe a grind that looks more like the early industrial age than a white‑collar revolution. One backend engineer in Seattle told reporters that teams moved from measured sprints to 996 and beyond; AI generates a flood of code, but humans must constantly prompt, review and validate it. Managers, increasingly dependent on AI metrics, reportedly expect “AI‑only” code paths and use token or PR volume as de facto performance indicators. That pressure is producing frantic reskilling — learn AI tooling or face being first in line for cuts — and spreading anxiety among parents and mid‑career staff who cannot easily pivot.

Corporate calculus and geopolitics

Companies argue they are chasing efficiency and the compute race. It has been reported that some firms, while courting clients such as OpenAI, are also using AI outputs to justify eliminating roles like technical writers and other supporting functions. This dynamic plays out against a backdrop of intensified U.S.–China competition over AI infrastructure, chips and cloud capacity, where firms face incentives to scale compute while trimming labor costs. The mix of booming profits, aggressive datacenter buildouts and mass layoffs raises broader questions: is this technological progress — or a new form of workplace extraction?

What happens next?

For now, the industry is caught between a productivity metric that rewards raw output and a human workforce scrambling to keep up. Some veterans shrug it off; others are leaving or retraining. Reportedly, internal “AI Weeks” meant to boost skills are also being used as efficiency audits to guide cuts. The net effect is clear: faster code delivery, larger codebases, and a tighter cage around the people who must still make sense of it all.