Can you tinker your way out of the permanent underclass?

Kelsey Piper with the compelling argument that escaping the "permanent underclass", if that is indeed a possible future, is best done through politics and not through getting really good at today's AI tools.

More pragmatically, her experiences match my own: always use “extended” or “high” thinking modes with every model (the one that’s one step down from their ‘research’ grade thinking level that tends to produce long reports) and stop writing really long prompts. The time you’ll save is significant and your outputs will be better.