So this is more an autocad equivalent? Does it have interfaces to enable reusability in some other file format (e.g. any way to get it into generic models eventually importable into revit?)

I assume they're saying blow up the hardware marked so they can liquidate assets efficiently?

Theres some argument they might be 208 (75% power for resistive heat), rennovations for apartments suck, etc.

The play store doesn't even update my apps for me anymore often

I don't think that's true that they use them because they're linked to slot machines. Sometimes variance just adds variety, which is valuabke in itself.

I also don't agree if you're implying this optimization is intentional. It's often just a way to make the game longer. People don't like a game they complete quickly

in house chips are pretty cool. Shared memory is a really nice architectural advancement.

What is your specific definition of "in house" chips ?

Usually we'd apply that to Samsung's exynos or Sony's image sensors for their DSLRs. Would Google's Tensor for instance fit in that definition ?

I thought they're buying a custom model they can run (likely for privacy reasons)

I don't want to accuse this of being AI generated, but what's with some of the graph choices?

The behavioral changes one seems very unreadable.

The last post was similar: https://news.ycombinator.com/item?id=45899331

That graph is not very readable, you are right. Since I don't have a unit to measure the dog's behavior changes, this is reflected in the graph

Your solution to not having a unit is graphing it with an unreadable scale???

Looking at the author's other posts: Very prolific, long articles, and weird outliers like https://lightcapai.medium.com/should-i-go-milan-211ef725bb71, https://lightcapai.medium.com/if-you-are-suicidal-read-this-..., and https://lightcapai.medium.com/alpay-algebra-solving-the-unso... (a weird take on the halting problem).

Yeah, they're mostly fluff. Based on the title I'm a bit disappointed there wasn't more actual science here, like measuring bacterial content changes in the stool.

While I don't doubt the perceived changes and there is plenty of human research indicating that gut bacteria dictates more about us than we'd like to admit the graphs in this article definitely portray the results as having more rigor than they actually did.

Real time learning that doesn't pollute limited context windows.

You can mimic this already. Unreliable and computationally inefficient, but those are not fundamental limitations.

Does the future productivity growth that would have been gained later (due to more junior engineers not entering the field) outweigh the AI gains?

If it's just the little productivity boost now, I think it's a net negative if hiring trends continue.

I think it's a discussion to be had but talent pool is a tragedy of the commons situation.