Zoom out take on AI state of play (+ therapeutic cope) My biggest fear remains whether enough land / power / shells get built to absorb compute supply. The bearish-bullish case, if you will... near-term numbers miss for some, but extended cycle yada yada. Current guess is there's enough, but I say this humbly. Rest of world can absorb more than people think (Latam is the new hot topic), but the builds are complex and the clock is ticking. However, last night's ANT news and today's TSMC news make me more optimistic. 1) TSMC is very conservative and doesn't freelance. They're sending teams on the ground to make sure the DCs aren't fake news... and they're still raising capex / sounding confident. I take that more seriously than non-industry experts tracking satellites, etc. Sorry if that offends 2) Frontier lab economics today make more sense than I thought. Financial reporting / readouts on X suggest solid margins even after training costs, with positive FCF (for now). Wow. And that's YTD, including 2 less scaled, lower-margin quarters + paying Elon bananas rates. (Too bad Elon isn't as generous as he'd like Hynix to be...). Compute costs should improve as non-Elon capacity comes on. And we'll soon get beyond safety / distillation headwinds. 3) The implied ARR #s# are lower than many thought (though oddly in-line w/ the analysts I speak to most). Gross/net confusion, bad reporting, fake news... exhausting stuff generally. BUT imagine how different people's tunes would be at $85bn ARR. It may be there now, given the compute coming on and excitement around the 5.5 family showing up in public data. Or it could get there in a few weeks. So a 4-6 week delay to the target number and it's all over? If that terrifies you, you should have been way more terrified at nearly every step of the last 2yrs. Particularly in June!! BTW *nobody* back then (me included) envisioned them eating the safety, the MSFT/Meta, and distillation hits all at once. Anyway, I expect a vibe shift quite soon.