Why the AI job loss narrative is overhyped: AI models still need to be prompted and verified, often iteratively, to drive business value. As @balajis says, AI is middle-to-middle, not end-to-end. Humans do the stuff at the ends (supervision); AI does the stuff in the middle.

"AI is amplified intelligence, not artificial intelligence...That just means the smarter you are, the smarter the AI is." ~@balajis

Important blog post by @balajis rebutting the prevailing narratives around AI job loss and the future of AI being a centralized superintelligence which controls us.

Ten thoughts on AI.

The beginning of this pod is so good. @balajis and @martin_casado deconstruct the AGI/ASI mythos, exposing how those myths are based on unphysical assumptions The most advanced AI provably still couldn't predict chaotic systems or break hashes.

Prompt your AI like the old Twitter. Limit responses to 140 characters, or 140 words, or perhaps 140 lines. Otherwise you’ll get back 140 pages.

The way Bitcoin scales is offchain and other chains. Practically speaking, you either load BTC into an exchange and then do offchain transactions, or use something like wrapped BTC on another chain. That’s more hub-and-spoke than fully p2p, but anyone with BTC can set up a hub.

.@balajis says every startup now needs a founding creator. Someone who speaks for the product before it ships. "The founding engineer is the how... The founding creator is the why." If you're outsourcing content, you're outsourcing your edge.

In light of the hit piece on Luke Farritor, @balajis returned to the podcast to give a love letter to legacy media in only the way he can—and to provide a new path forward for tech.

This actually makes sense to me, because AI is great for images, video, user interfaces, and anything where we can verify by eye. And Replit generates useful prototypes whose functionality can be instantly visually verified. It makes easy anything that should be easy.