It's worth listening to what Karpathy actually says, not what people are saying he said. It seems like his views are fairly similar to yours about this being a continuation of automation progress. As to the AGI in 10 years? When Dwarkesh kept trying to pin him down about "AGI when", he said "I almost refuse to answer that question."
What he actually said is that this is going to be the decade of agents, that the technology is not through yet. In particular that he is not using agentic coding for his current coding task which is very peculiar. For nanogpt he is trying to write unusually clear code for educational purposes, and he complained that the coding agents kept wanting to write it the ordinary way. So he is using the cursor style completion mode.
He doesn't expect any sharp inflections in perceptions of AI capability or economic indicators as the technology develops.
It's worth listening to what Karpathy actually says, not what people are saying he said. It seems like his views are fairly similar to yours about this being a continuation of automation progress. As to the AGI in 10 years? When Dwarkesh kept trying to pin him down about "AGI when", he said "I almost refuse to answer that question."
What he actually said is that this is going to be the decade of agents, that the technology is not through yet. In particular that he is not using agentic coding for his current coding task which is very peculiar. For nanogpt he is trying to write unusually clear code for educational purposes, and he complained that the coding agents kept wanting to write it the ordinary way. So he is using the cursor style completion mode.
He doesn't expect any sharp inflections in perceptions of AI capability or economic indicators as the technology develops.