iamlemec

Thanks, looks really interesting. Do functions like matmul support inputs of differing type, like say A=int8 and B=float? Would be nice if you could get memory efficient quantized matmul with operator fusion.

I believe `count` will only give you the number of non-null rows, so the numbers from the first command could differ by column if there were null values. You can also use the `size` command to get the total number of rows, and that will return a `pd.Series` with or without a column specifier.

Just another data point here: I'm running a relatively high-end workstation on Linux/Fedora 36, and it is quite smooth in Chrome but somewhat sluggish in Firefox.

Oh that looks excellent, I'll check it out. Yeah, the latency right now is decent enough for whiteboard writing but could definitely be better. Only tricky thing might be that some apps have AirPlay integrations that will just broadcast parts of the screen, like only the writing surface without the UI elements up top. Not sure if that carries over to NDI, but I could always just crop it in OBS.

Ah yeah, I forgot to mention that I'm on linux, and iPad direct to Zoom doesn't appear to work there. Also, going through OBS lets me do scene switching between full video, slides, and whiteboard, the latter two with video overlay in the corner.

This has been really useful for me when teaching online. Just mirror my iPad to a desktop window, then point OBS at said window (then pipe OBS to Zoom through a virtual camera). I tried to modify the source to output to a v2l loopback device instead, but couldn't quite get it to work.

You can add custom CSS in Preferences → Appearance. I've set mine up to look something like the mobile version for a while now. I'm always a little shocked seeing the original when logged out or on someone else's computer.

The cell data is always stored as plain text, on the client side simply with an attribute. And you can only ever edit cells as text, so those issues never come up. I could see how the checkbox would be tough though because you'd have to decide which line they clicked on in the text representation.

I've thought about porting to Jupyter too, but the plugin system on JupyterLab is pretty involved, and you lose plain text storage on the server side.

I've made a web-based one called Elltwo where you can toggle between plain text and rendered markdown on a line-by-line basis. I've found it to be a good midpoint between full on WYSIWYG and just editing raw text. It also supports equations, references, footnotes, etc.

Demo here: http://dohan.io

It was on STS-27. It's mentioned in the Wikipedia page for that mission and also in this article:

https://www.americaspace.com/2018/12/09/dying-all-tensed-up-...

Definitely consistent with my experience. One time I was in a restaurant where this sort of arms race dynamic was occurring, and at some point a guy just did one of those really loud two-fingered whistles, then in the resulting silence said something like "let's all try to keep it down in here". It worked incredibly well, and the effect was actually very persistent.

The precondition here is, of course, incorrect. The latest Pew survey puts the percentage of Republicans who believe in anthropogenic climate change at about 25%.

http://www.pewresearch.org/fact-tank/2018/05/14/many-republi...

Thanks! Just updated the demo document to highlight the most important commands.

Direct link to live demo: http://dohan.io