@abucci I see what you are saying. My thought is more than with agentic harnesses, speed isn't the edge that has been gained to combat the thesis.Instead, we are increasingly able to chunk out pieces of code, have multiple agents run 24/7 validation loops, run 100s or 1000s of tests on tweaks or adjustments etc. Humans can be in that loop. The sheer coverage has not been possible before, and while I take your point that LLMs might be inherently less capable that a skilled programmer, the scale and attention being applied can already make up for some of this (and over time, potentially surpass it).For me, its clear the economics don't yet work for the above, but we are marching towards it.I've found myself thinking of the infinite monkeys with a typewriter will eventually write Shakespeare....
