Meta’s 75% rule: Why your next mobile update will be

The human software engineer is officially an endangered species, thanks to the Meta 75% rule. If you thought the Year of Efficiency was just a corporate buzzword to appease Wall Street, think again. According to internal documents surfaced in late March 2026, Meta has stopped merely suggesting AI adoption and has started mandating it. We are no longer looking at a future where machines help people code; we are entering an era where people are simply the janitors for the machines.

What Happened

• The 75% Mandate: Quality is No Longer the Metric
• The £105 (about $135) Billion Gamble on Personal Superintelligence
• Why This Should Terrify Samsung and Apple
• The Death of the Craft and the Birth of the Prompt

The headline figure is staggering: Meta’s Creation Org, the massive team responsible for the very apps you use every hour, including WhatsApp, Instagram, and Facebook, has been handed a directive that 65% of its engineers must write at least 75% of their committed code using AI tools like Metamate and Google’s Gemini. This is not a pilot programme, it is a fundamental rewiring of how software is built, and it is happening while the company simultaneously clears out hundreds of “legacy” employees who do not fit the new AI-native mould.

The Meta 75% Rule: Quality is No Longer the Metric

For decades, the gold standard of software engineering was craft. You hired the best minds, paid them handsomely, and expected elegant, efficient, and secure code. As People Matters reports from the leaked targets, that world shifted this week. By enforcing the Meta 75% rule for AI-generated code, Mark Zuckerberg is effectively telling his staff that volume and “agentic impact” matter more than human oversight. According to our news coverage of the leak, the Scalable Machine Learning team has gone even further, aiming for up to 80% AI-assisted code by the end of this half.

The £105 (about $135) Billion Gamble on Personal Superintelligence

You have to wonder where the money is going. Meta’s projected capital expenditure for 2026 is a mind-numbing $115 billion to $135 billion (around £92 billion to £108 billion). Most of that is not going to people; it is going to Nvidia GPUs and custom MTIA silicon, as Engadget has tracked in its Zuckerberg-on-AI coverage. As we have noted in our editorials, Meta is now spending more on electricity and chips than most sovereign nations spend on their entire militaries. To fund this, the human headcount must be sacrificed. The redundancies hitting Reality Labs and the sales teams this week are just the beginning of a much broader restructure.

Why This Should Terrify Samsung and Apple

The danger for the mobile ecosystem is a sudden, sharp decline in OS-level stability. As we explore in our AI coverage, the more we rely on agent-assisted development, the more we lose the institutional knowledge of how these systems actually work. When the AI writes 75% of the code, and the AI Builder only understands 25% of it, who fixes the bug that only appears once every million cycles? The machine cannot fix what it does not truly understand, and the human has forgotten how to read the language of the machine.

The Death of the Craft and the Birth of the Prompt

This is not just a change in how Meta works; it is an admission that they no longer value the people who built the company. By rebranding engineers as AI Builders and forcing them to hit arbitrary machine-usage targets, Meta has turned its workforce into a glorified QA department for an algorithm. The Year of AI Autonomy is here, but it feels less like progress and more like a surrender. We are watching the slow-motion shift of engineering as a profession, replaced by a high-speed assembly line of automated mediocrity.

Ultimately, Meta’s 75% rule is a confession of where Zuckerberg thinks the future is going. It is an admission that they want to innovate through algorithmic brute force as much as through human brilliance. If your next mobile update feels like it was written by a machine, that is because much of it was, and the person who built it may have been too focused on hitting their AI-usage quota to notice the seams. Zuckerberg is chasing personal superintelligence, and we are the ones who will have to live with the results.

What Meta’s 75% AI-code rule actually changes for the apps you use

Meta’s 75% AI-code rule is a directive for engineers, not a release-gate, and the immediate effect on your mobile apps is subtle. Feature velocity in WhatsApp, Instagram and Facebook is likely to increase in the short term as AI-written pull requests ship faster. The medium-term risk is regression: AI-generated code performs well on boilerplate and narrow refactors, and less well on the kind of rare-path concurrency bug that only appears at Meta’s scale.

The internal playbook behind the 75% rule is also a cost story. Meta’s 2026 capex is skewing so heavily toward GPUs and data centres that the operating budget can no longer afford unrestrained human-engineering growth, as Blockchain Council’s coverage of the mandate notes. Pushing 75% of code through AI tools is partly a productivity bet and partly a headcount hedge; both of Mark Zuckerberg’s options converge on fewer humans shipping more features.

For users, the honest expectation is a slightly more glitchy release cadence over the next six months while Meta calibrates what AI-written code can and cannot safely own, followed by a harder-to-call long-run equilibrium. If Meta’s tooling is as good as Zuckerberg claims, users will never notice. If it is not, the bug reports will get noisier before the process stabilises, and the “AI-native” pitch will look less like an inevitability and more like an expensive experiment.