bensandcastle

This is a mainstream media piece. Technical writeup and papers will be forthcoming.

The scaling issue the author identifies is exactly what we've built Strong Compute to solve. Dynamically scalable super computing.

"In this environment the problem with a single big unit is that significant updates to it are likely to be quite expensive and you probably simply won't have that money all in one lump at once."

This is true if you approach it as a long term lease, or a fixed purchase, which is often a requirement given even if using on demand or spot there's still a lot of setup work to do. Our focus to crush this issue. Zero code change from a small VM to a large cluster.

The natural light and diffuse light are good tips.

Next is to get a big screen eg. 85" 4K and put it 1.5m away. That should be your main display. I don't have that all the time, but then I get some variety, 85" @ 1.5m a lot of the time, laptop some of the time, driving/walking etc. for longer range.

1.5m is the midpoint of focus for the muscles in your eyes.

I built augmented reality displays and this was the focal plane we selected for to minimize eye strain and the felt sense of vergence/accommodation conflict.

We could then throw graphics as close as ~30cm, or at infinity using vergence adjustments, even though the accommodation was at a fixed 1.5m. Graphics felt best at that distance, but they also felt ok in the range 0.5-10m, which suited nearly all productivity scenarios.

marginally relevant. space based dawn dusk LEO solar infra is the answer. vastly more power than we'll ever get on the surface of this rock and then onto Sol.

self built runner box how to including overclocked tuned high spec CPUs, and how to deploy self hosted runners.

how much analysis have you done on this? I'm working on a series around Kardashev II work - space solar, asteroid mining, dyson spheres and the real work that is happening to support this now. contact me if you'd like to discuss - @bensand on X

Ah yes! I remember working on this. It was born out of asking all the YC founders what they wanted to see - and anyone else we met as well. Formed the basis of our second kickstarter which we self hosted and was much more successful.

Yes, correct. As in their degree of openness (not much) is what's letting them down. I see how you could have read it as I meant their too open, which I definitely don't think is the case!

a bunch at Vision Pro, some at Zuck's Meta, some at Hololens, some doing other things. Meron is doing BCI, I'm doing AI infra - strongcompute.com (YC W22)

Yes, got my founders ed. pre-order in, can't wait!