fzimmermann

The limit of 5 decimal places makes it a bit impractical for my use case: While debugging, I often use fast numerical approximations to obtain out some number, that I know should somehow relate to some properties of the input, combined with some constants. But the precision is often less than 5 decimal places. Example: The operator norm of some operation or some integral -- is it pi/4inputsize? ln(2)inputsize? sqrt(1/2)*inputsize?

https://www.wolframalpha.com/ is really useful in that case

Does anybody know if this would also work for tablets? It has been quite some time since I tried last, but you used to have to use the web app to check your messages on your tablet. Would be nice if both devices could now share an account

Reading the parent in the most favorable light: "Attention is all you need" can be seen as a fundamental paper for the current LLM generation, which indeed came out of google brain..

TIL: There are replaceable SMS which you can update after they have been sent. I wonder why not every 2FA service is using those?

Do you remember which extension?

Yes, it would help if those key combinations would work (even though they are still at least unusual). The quit combination does not work and neither does evaluating a cell. Does anybody have luck actually getting it to work?

Edit: The issue is using a new environment without ipykernel installed (not a dependency of the conda package). With ipykernel, I can exit nbterm and even evaluate cells

Ok, I failed in basic usage: Entering nbterm, I assumed some standard key bindings -- None seemed to work. Neither Ctrl-C, Ctrl-D, Ctrl-x ctrl-c, escape q, escape :q, ... seems to work. So off to google: Escape Ctrl-q should work. But does not.

Ok I'm stuck, let's at least get 1+1 to work. So entering 1+1 and .. hitting Enter? Shift Enter? No. Ok, command mode, Ctrl E.. aaaaand it's stuck.

Looks nice but I just don't seem to get it to work using anaconda on macos..

I see some issues with the problem description and the solution: The author assumes the problem can be split in x and y independently. Firstly, this has not to be true for all projections and grids. Secondly, the solution minimizes Manhattan distance in lat/long., not actual distance (you can construct points where the minima are different). This will not matter if you can approximate both grids as regular (equidistant and rectangular grids). But in that case, you could easily calculate for each point in one grid the corrosponding point in the other grid by scaling and rounding...