Seems like an update towards exponetial growth potentially being robust. However, we’ve recently seen that exponetial growth on software task length may not translate to real software engineering usefulness.
Also, this does not necessarily imply exponetial growth in agency, because agency depends on unlocking increasingly hard bottlenecks. The growth rate is much slower for self driving, while other agentic tasks have a fast growth rate but are far behind. That means that overall agency might turn out to grow sub exponentially, even if growth is exponetial at every individual task.
Interesting.
Seems like an update towards exponetial growth potentially being robust. However, we’ve recently seen that exponetial growth on software task length may not translate to real software engineering usefulness.
Also, this does not necessarily imply exponetial growth in agency, because agency depends on unlocking increasingly hard bottlenecks. The growth rate is much slower for self driving, while other agentic tasks have a fast growth rate but are far behind. That means that overall agency might turn out to grow sub exponentially, even if growth is exponetial at every individual task.