Andrew Bosworth thinks AI gains should mean more output rather

The Meta CTO just made it clear that the productivity boost from AI isn't meant to be a ticket to a four-day work week. In his view, if an LLM agent helps you finish a task in two hours that used to take eight, the expectation isn't that you log off early—it's that you find six more hours of high-value work to tackle. It's a classic "efficiency paradox" where the more productive we become, the higher the ceiling for what's expected of us.

For those of us experimenting with a real-world AI workflow, this is a polarizing take. On one hand, it makes sense for a hyper-growth company. If you can suddenly 10x your output, the competitive advantage comes from doing 10x more, not just doing the same amount of work faster. But for the average developer or prompt engineer, there's a real risk of burnout if the reward for efficiency is simply a heavier workload.

How this actually looks in a dev workflow

If we apply this logic to a practical tutorial on increasing throughput, the "Meta approach" would look like this:

1. Automate the boilerplate: Use an LLM to scaffold your API endpoints or write the initial unit tests.
2. Rapid Prototyping: Instead of spending a day on a technical design document, use an AI to generate three different architectural options and spend your time auditing them.
3. Immediate Refactoring: Use the time saved to go back through the legacy codebase and fix those "TODO" comments you've been ignoring for six months.

Basically, instead of the AI being a tool for leisure, it becomes a tool for expanding the scope of your role.

• The Old Way: Spend 40 hours on 2 features.
• The AI Way: Spend 40 hours on 8 features because the "grunt work" is automated.

This shift puts a massive premium on prompt engineering and the ability to manage AI agents. If your boss expects more output because you're using AI, your value is no longer in "doing the work," but in your ability to orchestrate the AI to produce high-quality results without introducing hallucinations or technical debt.

I wonder if this creates a "productivity treadmill" where we're just running faster to stay in the same place. If every company adopts this mindset, the baseline for a "standard" workload will just shift upward. We're moving from a world where we are valued for our labor to a world where we are valued for our ability to direct compute. It's a weird transition, but it's clearly where the big tech firms are heading.