computer,
I've done a 180 on AI in the last month.
Not all the way in. I'm not with the AGI singularity nutcases. I've read Stochastic Parrots. I've read Iris van Rooij. I don't think AI is getting to general intelligence anytime soon — we can't even define human-level intelligence, or intelligence at all, in any rigorous sense.
but it's useful now.
Last spring, I tried to use a frontier-level AI agent to add a simple feature to a C++ codebase. it failed hard. confidently insisted it had solved a problem that was very obviously not solved. then at the end of 2025, I tried JetBrains Junie as an experiment. night and day. it built what I asked for, credibly, in one shot. it needed refinement, but it was a functioning program that did what I asked. I've basically not written code since.
AI isn't solving world hunger. but, if well guided, it can solve almost any problem solveable with software.
I think it's going to change the trajectory of humanity roughly the way the PC did. The PC took computing off the mainframe and put it on your desk — but you still needed expertise, still needed money. AI could unlock the Star Trek model. computer, — your computer, your problems, but now it understands you. not thinking for you. just an interface.
That's what PARC was. the mouse, the window, the desktop — not a new kind of intelligence, just a better way to talk to the machine. AI is that, again.
I don't need a superintelligence. I need my computer to know that 99% of my inbox is bullshit and help me separate out the 1% I want to keep.