But I’m responding to the rescue mission comment, which, since Vietnam, have overwhelmingly employed helicopters (Huey’s then, Black Hawks today). But machinery aside, the larger point is that air operations will likely go worse here than they did in Vietnam, unfortunately for both sides.

Here’s a use case that may illuminate the difference, from my own work at Nvidia. Im currently training some large sparse autoencoders, and there are issues with dead latents. Several solutions exit to help here, such as auxk, which I can certainly include and tune the relevant params as you describe. However, I have several other ideas that are much different, each of which requires editing core code (full evaluation changes, initialization strategies, architecture changes, etc.), including changes to parallelism strategies in the multi-rank environment I’m using. Moreover, based on my ideas and other existing literature, Claude can try a number of new ideas, each potentially involving more code changes.

This automated run-and-discover process is far beyond what’s possible with hyperparam search.

But these are all additional points in support of what is already said.

“General and administrative expenses increased by $68.1m ... The increase was primarily driven by … an in-person company event held in Q3 2025”

From New Yorker profile: “His goal… is… by making information freer, he hopes to make the world fairer, kinder, and nicer.”

Where he also writes, “I definitely feel the most fundamental issue is economic equality.”

But hey, the stock is up 25%!

Also: plenty of meaning outside of running a SaaS. Hell, undergraduate research assistants probably contribute more to societal at large.

Here’s why what you wrote seems needlessly contrarian: Amazon just posted an $18B quarter, so there is no pressing financial pressure. Okay, so you suggest this may be a last resort in lieu of retraining, but we’re talking about 14k jobs across many teams (I know of at least 40 affected), levels, and job families. The idea of needing to cross train is obviously not the culprit at that scale; An SDE laid off from one team can easily perform the same tasks on many others internally. This also completely ignores how Amazon works internally, with managers required to rank employees for pip, and, for events just like this one, URA, regardless of whether or not they deem them to be competent or not.

Of course, Amazon has also been documented to use automated processes for pip/layoffs, and the idea that layoffs involved any ounce of consideration as a last resort is so unbelievable it feels almost inflammatory.

The notion that criticizing one of history’s most profitable companies laying off thousands (at the height of their profits) is the same thing as stating, “every company beyond profit X should never do layoffs” is a blatant misrepresentation and ignores any context.