How is a data center defined? Is a small business dedicating a room to server racks a "data center"?

The law defines it as (approximately) anything connected to a public utility that has more than one megawatt of computing power.

With the change to only five days of being publicly traded requirements, incentivising market makers to keep a high valuation becomes very cheap.

After five days the index funds have to buy at the last price making it final.

In other words even if the vast majority of the market believes it's worth much less, they can force a high price and force basically everyone to hold it via retirement funds.

Yes exactly what I said - you can take issue with how it was included, but if inclusion == force then you're also forced to buy Dell, Apple etc.

But Dell Apple etc were included when stronger requirements were in place.

Comforting the passive investor that the market would have had time to react and to force a lower valuation before their inclusion in the index

Yes they should have stick with the naming convention.

google/gemma-4-31B-it-ass

is it still a useless model if, say, you can run it at (prompt+output)*24/s and use it to make executive function decisions?

Mmh yes and no. There are two arms in this hypothesis and you can summarize the second one (the main one) as basically wanting to piss off / punish the Europeans. This is perfectly on point and fits the character.

Occam's razor might well describe actions of a cold calculating leader. But of Trump..?

Interesting. When it's the state I think the overwhelming opinion is that predictive policing is dangerous but when it's a private company we actually want it to enforce it?

They could not be held accountable to warn her if they had not done the analysis. They did. Their organizational conclusion was that it was potentially an unsafe trip. Shit, they could have just cancelled the ride dynamically and re-assigned her. Why wouldn’t they do that? It’d probably be more expensive. Maybe they’d get more cancelled rides. Maybe this woman wouldn’t have been raped by an agent of Uber selected for and sent to her by them.

Wouldn't they then expose themselves to discrimination and loss of revenue lawsuits from targeted drivers?

It depends. Are the inputs to the algorithm themselves discriminatory? If so, then yes that would be appropriate. But that is a different conversation. They determined the passenger may be unsafe and did nothing.

Mind you, these companies work very hard for us to not know how they match A to B, usually so we don’t notice things like their disregard for safety.

The inputs wouldn’t even matter; the inputs could even be above reproach but if there were disparate impacts in terms of outcomes, the case for liability could be made.

Maybe, but they’re clearly liable for not using the information.

Aren't you just moving the problem a little bit further? If you can't trust it will implement carefully specified features, why would you believe it would properly review those?

It's hard to explain, but I've found LLMs to be significantly better in the "review" stage than the implementation stage.

So the LLM will do something and not catch at all that it did it badly. But the same LLM asked to review against the same starting requirement will catch the problem almost always

The missing thing in these tools is that automatic feedback loop between the two LLMs: one in review mode, one in implementation mode.

I've noticed this too and am wondering why this hasn't been baked into the popular agents yet. Or maybe it has and it just hasn't panned out?

Anecdotaly I think this is in Claude Code. It's pretty frequent to see it implement something, then declare it "forgot" a requirement and go back and alter or add to the implementation.

AFAICT this is already baked into the GitHub Copilot agent. I read its sessions pretty often and reviewing/testing after writing code is a standard part of its workflow almost every time. It's kind of wild seeing how diligent it is even with the most trivial of changes.

You have to dump the context window for the review to work good.

Exactly, his whole tirade felt extraordinarily far fetched, sketchy if not outright racist.

Well it's 2025, we've just spent the better half of the year discussing the bitter lesson. It seems clear solving more general problem is key to innovation.

Hardware is not like software. A general purpose humanoid cleaning robot will be superior to a robot vacuum but it will always cost an order of magnitude more. This is different from software where the cost exponentially decreases and you can do the computer in the cloud.

I'm not sure advancements in AI and advancements in vacuum cleaners are at similar stages in terms of R&D. I'd be very wary of trying to apply lessons from one to the other.

Alternatively:

  enclave, err := secret.GetEnclave()
  // err contains whether the platform doesn't support it
  enclave.Do(f)