caminante

IMHO, the parent's point ties to the Bitter Lesson.

Use the best tool available knowing that advances in compute dominate clever optimizations.

If models stopped leapfrogging each other, then it wouldn't hold. But it appears to hold for now.

Details aren't as sexy as the headline would lead one to believe.

Instead,

> For three years the pitch from the big AI model labs has been beautifully simple. Software engineers are the most expensive employees on Earth.

Same formula across country, culture, time and people keep getting sucked in.

Unless you delete before scrapers get it, it's stored on multiple databases.

Anecdotally, I have noticed the same in the past week. It might just be anecdotal or driven by a long context window.

In medicine, it's a solution for someone who wants liquidity (buy kids new home, help local dog shelter, add a new mistress) and doesn't want to retire, yet.

The problem is that the terms are custom and YMMV as an existing patient of said practice.

For the parent, an extra 1M umbrella is an extra $2-300/y and covers beyond auto!

* "More efficient, next gen coal plants"

* "Receiving capacity payments subsidies to stabilize grid"

* "We won't run these plants at full capacity"

When their carbon intensity still can't touch natural gas alternatives at scale, state owned enterprises are operating in a command economy, and they have no teeth to their carbon commitments.

They don't have to build new coal at 10:1 versus retirements to "replace aging coal" or stabilize the grid.

Coal is just too cheap and adds diversity,while nobody forces CN to enter climate agreements and looks the other way.

That doesn't make it righteous and is a common critique of their coal powered PV manufacturing base.

They're playing games with carbon commitments with arbitrary reduction goals from undefined peaks. We can't trust them.

Meanwhile, they're currently building >200GW of new coal plants.