< Back to the archive

Like what you see? Subscribe here and get it every week in your inbox!

Issue #180 - August 21, 2022

Here are the top threads of the week, happy reading!

Top comment by rswerve

College towns. Pick some states, and go visit towns with colleges in them. They'll have restaurants and groceries, and those faculty kids have to go to school somewhere. Internet should be passable, too. They won't be the cheapest small town option, but houses should come in well below a million bucks.

The one town you mention, Lancaster, PA, is a model of the type.

Top comment by alanfranz

My 2c (not exhaustive for what you want to do, probably):

1) Get some statistics/probability basics. It's full of people (you can see a lot of analyses on Kaggle) that "do machine learning" but make very silly mistakes (e.g. turn categorical data into a float and use it as a continuous variable when training a model).

2) take a look at traditional machine learning approaches. Nowadays you're swamped by DL (a lot of good suggestions on this thread, I won't chime in), and you miss the fact that, sometimes, a simple decision tree, or dimensionality reduction approaches (e.g. PCA or ICA) can yield an incredible value in a very short time on huge datasets.

I had written a fairly short post about it when I finished my georgia tech path https://www.franzoni.eu/machine-learning-a-sound-primer/

3) It can take a lot of time to become effective in ML, effective as in, what you _manually create_ is as effective as picking an existing trained model, fine tune it, and use it. This can be frustrating: low hanging fruits are pretty powerful and you don't need to understand a lot about ML algorithms to pick them up.

4) Consider MOOCs or online classes. I took Georgia Tech OMSCS, I can vouch for it and some classes force you to be a data scientist and read papers as well, and you can have "real world" recognition and discuss with your peers, which is useful!

Top comment by hoechst

Not really a script, but a `.ssh/config` to automatically deploy parts of my local cli environment to every server i connect to (if username and ip/hostname matches my rules).

On first connect to a server, this sync all the dotfiles i want to a remote host and on subsequent connects, it updates the dotfiles.

Idk if this is "special", but I haven't seen anyone else do this really and it beats for example ansible playbooks by being dead simple.

   Match Host 192.168.123.*,another-example.org,*.example.com User myusername,myotherusername
      ForwardAgent yes
      PermitLocalCommand yes
      LocalCommand rsync -L --exclude .netrwhist --exclude .git --exclude .config/iterm2/AppSupport/ --exclude .vim/bundle/youcompleteme/ -vRrlptze "ssh -o PermitLocalCommand=no" %d/./.screenrc %d/./.gitignore %d/./.bash_profile %d/./.ssh/git_ed25519.pub %d/./.ssh/authorized_keys %d/./.vimrc %d/./.zshrc %d/./.config/iterm2/ %d/./.vim/ %d/./bin/ %d/./.bash/ %r@%n:/home/%r

Top comment by mg

I don't know about Chrome, but in Firefox:

    Chrome now hides the bookmark bar by default
Not sure if it is the default, but my Firefox bookmark bar is always visible

    load common favorite news sites & blogs
    or load a window with all my productivity SaaS sites.
Shift+Click on a bookmark folder does that

    pick up where i left off on a research rabbit hole
Right-Click on one of the open tabs, click "select all tabs", right-click it again and click "Bookmark Tabs"

    i want it to be intuitive, efficient, and a prominent UX feature set
I'm a heavy user of bookmarks, and I am pretty happy with the state of bookmarks in Firefox.

What I do miss is a way to back them up from the command line. Firefox has a nice "Export Bookmarks to HTML" function, but it seems only available from the GUI, so there is no way to automatically backup the bookmarks in this nice format. I tried for a while to extract them from the SQLite DB FireFox stores them in, but the layout of that DB is pretty complex and bloated. Even after a bit of fiddling, I wasn't sure I really correctly got the data out.

Top comment by gniv

Sometimes you can get lucky and land an interesting job, but I think getting to do interesting work takes time (and luck). An anecdote from my past as an illustration: When I started my job at [FAANG] I was assigned a project. It was not mind-numbing, but not sexy either -- modernize the infrastructure while not messing up the site. Being new, I gave it my all and was rewarded a couple of years later by having the option to work on a very interesting project, used by tens of millions, with a clear user-visible outcome. Afterwards I usually had my choice of projects and teams inside that org.

Edit: I guess the actionable advice is to get into a company that is big enough to have _some_ interesting projects.

Top comment by scarecrowbob

About half of the 30 or so bands I have played with have been abandoned after a couple of gigs, or non at all.

Mostly I've only just played music for fun. It's much more fun (and certainly easier) playing at some friend's house every week for 18 months than it is to book performances, market the band, produce works for the public, etc.

I have a sticker on my truck that says "This Sticker Will Last Longer Than Your Band", and that's been true.

I don't regret learning to play the entirety of Dark Side of the Moon and performing it once. I know so much Grateful Dead. I've play dozens and dozens of original songs written by friends that won't ever get any attention outside of the 20 people who were paying attention at some bar.

I've also played with more commercially focused endeavors and have an idea of the horrors of trying to tour and manage the commercial aspects of music. It's not like most of these bands were any worse or could not have been marketed.

I've played with a lot of folks who have only ever been in one or two bands.

The thing I find interesting about all that is this point:

they usually vastly underestimate how much work it takes to market a band

and at the same time (I find this a little beautiful) they vastly underestimate the goodness of their own musical output.

Just because projects don't succeed commercially does not mean they aren't well made or that these projects are not worth doing in as things in themselves.

Top comment by dan-robertson

0. RDP is not as simple as sending draw commands. It has eg commands that correspond to scrolling regions of the screen to save on network use, a framerate limit (25fps I think) and allows some colour space reduction to reduce bandwidth too. I think it does a bunch of raster things (eg maybe caching floating windows like right-click menus). I think the windows server implementation can take advantage of information about the composition of the screen from windows.

1. There is some trade off of latency for bandwidth: it may take more time to figure out a small change to send over the network. Looking at api use from eg X may help with old apps that make small updates but more modern apps (or even modern fonts) which just render to gpu buffers and composite are less amenable to this. Very modern apps that use special apis to do lower latency scrolling/resize may be a little better.

3. A lot of the time for Linux the solution is to use ssh and terminal apps as they tend to make smaller updates and require less bandwidth. You can also try mosh to compensate for high latency connections. Text editors can work ok in terminals, especially fancy modern ones with eg mouse support. And for web things you can do set up a socks proxy over ssh which I think can work for a lot of apps which are really just web sites. So this may be part of the reason: fewer people see Remote Desktop as necessary.

4. I’ve had reasonable success with xrdp on the server and a windows client. One needed to select 32-but colour to get a better protocol version and turning off double-buffering in some apps (eg emacs) helped. But that was over a wired high bandwidth low latency connection.

Top comment by landhar

I’m working on an open-source social bookmarking site in Elixir that is API compatible with delicious/pinboard. Its named linkhut and it’s currently able to import your bookmarks from pinboard and browser exports.

The flagship instance is: https://ln.ht

The source code is hosted here: https://sr.ht/~mlb/linkhut/

The documentation: https://docs.linkhut.org/introduction.html

The one thing that I’m working on before releasing 1.0 is taking a snapshot at time of bookmark and index its contents to make it searchable (similar to pinboard’s feature).

Top comment by jaitaiwan

Seems like you’re experiencing burn-out and from what I’ve read and experienced, burn out is a function of emotional investment vs the emotional return.

So I see that you have two options both of which shouldn’t happen before first taking a small break: 1. Continue to work in the industry and reduce your emotional investment. Hint: great places for that are big corporates. 2. Continue to work while upskilling in another field or your current field depending on what you prefer.

Depression often comes with the temptation to catastrophise the situation, avoid that urge. Seek out others who can be objective and talk you through it, be prepared to hear their answers.

I started out front-end, did a lot of my own side projects to get backend stuff. PHP is pretty good despite the hate because not many people want to do it these days but a legit and easy way to get some backend experience.

Top comment by jpgvm

By and large yes, people returned to the office however a lot are disgruntled and exceptions to work remotely largely depend on your manager and their ability to push that up the chain as Elon is meant to approve each exception personally. (in practice most of the full-time remote employees already had such exceptions but some newer hires during the pandemic weren't hired under that regime).

However a lot of folks quit but it has less to do with the remote work policy and more to do with a massive options cliff that occurred recently. Once the golden handcuffs came off a lot of folk figured it was time to move on, which is fair given the hell they went through to deliver 3 and Y + the 10x appreciation in their options.