A recent analysis offers a troubling historical pattern. Each technological leap, from steam power to the internet, arrived more slowly than predicted. Yet each arrival granted society a crucial grace period—decades, even—to establish safety laws and social norms before the technology matured.

That buffer is now collapsing. Research from the group Meter indicates AI task completion capability is doubling every seven months. Systems that in early 2024 managed minutes of work now autonomously handle complex projects spanning five to ten hours. If this trajectory holds, AI operating independently for days or weeks could emerge in the very near future.

History’s timeline is sobering. It took roughly 20 years from the internet's rise to implement foundational regulations. We had the same span between nuclear weapons and arms control treaties, a period filled with tense but contained conflicts that taught global leaders the boundaries. That learning curve may be a luxury we no longer have.

The transition from AI requiring close supervision to AI operating independently could be startlingly brief. In practical terms, the coding assistants I used six months ago needed constant correction. Today, tools like Cursor and Verdant execute intricate tasks with minimal intervention. The capability jump within a single year is unmistakable.

This acceleration creates a profound mismatch. Institutional trust, regulatory consensus, and international frameworks are slow to develop. They cannot be trained on a seven-month cycle. The central question is no longer if we will build guardrails for advanced AI, but whether we can possibly build them in time. The historical playbook, it seems, is obsolete.

Source: Reddit AI