In my case I started with something somewhat like Playwright, and claude had a habit of interacting with the app more directly than a user would be able to and so not spotting problems because of it. Forcing it to interact by pressing keys rather than delving into the dom or executing random javascript helped. In particular I wanted to be able to chat with it as it tried things interactively. This is more to help with manual tests or exploratory testing rather than classic automated testing.

My current app is a desktop app, so playwright isn't as applicable.

I don't have any experience in this area, but it seems like for every square meter of solar panel you need about half that in radiator area. And depending on your orbit, these are probably not static things just sitting there, they need to be orientated correctly to work and their correct orientations will change over time.

The worry for me is the level of human maintenance required. The ISS has probably the biggest solar array around, and they send humans out to perform maintainance and repair on it multiple times a year. A decent size data center would need an order of magnitude more solar and radiators than the ISS, and so presumably would need even more maintenance.

I got it to build an MCP server into the app that supported sending commands to allow Claude to interact with it as if it was a user, including keypresses and grabbing screenshots, and the difference was immediate and really beneficial.

Visual issues were previously one of the things it would tend to struggle with.

As far as I can tell from random articles online, it seems that as a rule of thumb, you need about 6 humans +1.5 humans per megawatt - and that's just for running the datacenter part, different people maintain the power generation infrastructure. Now, if you have to house those people in space or fly them up whenever they have to do anything, that's going to destroy your budget.

If you want to assume a level of automation that makes that unnecessary, that's fine, but then you need to also assume that same level of automation in earth based data centers too, and everything that goes with that.

By the way, 10,000x inference growth would look like what happened with cryptocurrency mining - after a couple of years, you'd be needing to upgrade all your machines with ASICs and the market would be flooded with very cheap graphics cards. I doubt that upgrading space data centres would be fun.

I say this tongue in cheek, but in all seriousness, I can't really think of any other benefit, and I no longer have a lot of faith in the good sense of some of the people involved.

My standard test for that was "Who ends up with Bilbo's buttons?"

Phones are never going to run the largest models locally because they just don't have the size, but we're seeing improvements in capability at small sizes over time that mean that you can run a model on your phone now that would have required hundreds of billions of parameters less than 6 years ago.

Every month, the thesis seems more certain that a mathematical model will be able to produce output indistinguishable from the output produced by a conscious, biological creature.

Once that's accepted, then the only interesting questions that are left are by definition unobservable. Is there anything that it is like to be that mathematical model?

The worst show I've seen for this was american - mythbusters.

The Tchaikovsky novella I really like is Elder Race. Technology-as-magic is done in so many places (Ventus is another favourite), and I usually enjoy it, but I felt that in Elder Race it was pulled off in an unusually elegant way.