UberMouse

>On the 20x plan, 1% of weekly limit is worth about $20 in API prices. A full week is then about $2,000

Considering I use on average close to $3k a week at API prices on my 20x plan I don't think this quite accurate. Some weeks I've used >$3.5k

And lets not forget the `claude agents` view and the backgrounding system. A++

The Workflow engine, the Mod system. I've tried to explore other coding harnesses and nothing is as powerful as claude code (barring investing a lot of effort into custom plugins for Pi or something). Other harnesses have workflow engines, but none are as powerful as Claude Codes.

Honestly that seems like a good result to me? Did you look at the product? Earbuds Connected by wires to a thing that you can dangle around your neck

5090 was an example, same process applies to lower tier GPUs that don't require extra power cables. ie a 3080 with the same power budget as a 1080 would run circles around it (1080 with default max power limit of 180w gets approx 7000 in TimeSpy, 3090 limited to 150w gets approx 11500). Limiting the power budget is very simple with tools such as MSI Afterburner and others in the same space.

Then get a modern GPU and limit the power to what your 1080 draws. It will still be significantly faster. GPU power is out of control these days, if you knock 10% off the power budget you generally only lose a few percentage of performance.

Cutting the 5090 down from 575w to 400w is a 10% perf decrease.

Brandon Sanderson only wrote the last book though? (which he split in two because all the stuff Robert Jordan wanted to cram in would have been a ridiculous 2000+ page tome) so it was already going to be at least 11 books by Robert Jordan. FWIW the two Sanderson books are my favourite of the series (but they are also the ending so it's hard to say)

That is talking about the 2023 Executive Order about training AI. Not the recent one about chip exports.

That would be true for the 4090, but the 3090 is even more cut down so the 5090 is about 100% faster in rasterization (based on the 4090 being nearly 70% faster than the 3090 and the 5090 being approx 25% faster than the 4090)

`bar` is 2 because `setFoo` does not update the value of `foo` until the next render

And then you go full circle when the project is bigger and Find Usages no longer works across projects (specifically in the context of a typescript monorepo) and it's back to grepping for usages.

Just chipping in as someone who is a tremendous fan of the books (I've read all of them 4 or 5 times and listened to the audio books twice). I _hated_ the series, I was very excited for it, watched Season 1, was blown away by how awful the changes were and won't be watching more.

It honestly baffles me that anyone who liked the books is able to enjoy the TV show, it feels like bad fan fiction.

Enabling the `envfs` service fixes the bash shebang issue `services.envfs.enable = true;`

Disclosure: I don't know if this has any downsides, I haven't run into any

I could hit triple digits WPM without knowing how to touch type. I just learned adhoc how to type very fast from gaming as a teenager. I haven since learned how to touch type (and moved to a split keyboard with Colemak-DH) and I type at a pretty similar WPM, it's just more comfortable.

Not if you have a keyboard with easily reflashable and programmable firmware (such as the Moonlander). If the firmware has built in support for layers it becomes trivial as you can just have a QWERTY layer activate when you hold down ctrl etc (Moonlander also)