This is what (pre-IPO) peak bullshit looks like.
I'm really pro-AI and I'm very excited about future developments, but the catnip is clouding our judgement.
Anthropic also say we should "pause" because we will "lose control" of a technology that has arguably less autonomy (now, in practice) than hand-written Python code. We are not "doing more with less" with current AI (if you take errors, legibility, time, cost and tail risk into consideration). Fundamentally new research is required, we are not even close.
There is weak evidence (so far) for broad economy-wide displacement or a general AI-driven "unemployment shock". There is some (possibly transient) suppression of junior hires and contractors (for highly automatable jobs).
They are now convincing Trump that nationalisation is a "good idea" for regular folks about to be "displaced" by AI.
> “It almost becomes a partnership with the American public,” he said. Trump added that “the American people can benefit from the success of AI, and by that, they’re going to like it better.” [WSJ]
George Carlin did say that "bullshit was the soundtrack of America". Imagine what he would have to say about all this if he were still alive.
This might actually be the most sophisticated, multi-layered bullshit in human history.
Book volumes, Netflix shows and management modules will be written about the 2020s frothy AI psychosis pandemic.
And what makes it all the more annoying is when you strip away the bullshit and see it for what it is: current AI genuinely is innovative and useful, but its harms can only be mitigated with high levels of AI literacy.
I'm hopeful that strong engineers will figure this out collectively, but I estimate it takes, on average, 18 months to 2 years of daily experience with vibe coding to properly grok it. We might not have that long.
PS. Look at the leaked Claude Code source
