What are the best secrets of great programmers?

The best secret I learned from a great programmer: Pay to Win.

It’s also good to Play to Win (which means building your own tools and libraries to accelerate your work), but committing to paying in the right places.

The first lesson I learned was how he used to spend his afternoons working from a Starbucks via a remote desktop to his office machine that would then compile code via a server farm. This was in the late 00s, but even then the connection was good enough.

Firstly, the bare minimum — I’m subscribed to:

1. Jetbrains (for access to the very best IDEs available).
2. Adobe (so that I can open Illustrator and Photoshop files if I ever need to, and whenever I need to edit videos).
3. Spotify and TIDAL (for the best in background ambiance). Not specific to programming, but thought I’d mention it.
4. Medium (it pays writers and gets me access to the best Data Science articles).
5. HP Instant Ink (you’d be surprised how much more you print convenient things when you don’t care about running out of ink).
6. This one is really easy to miss. Think of all the typos you’ve found in your CV months later when you saw it printed out or documents that just don’t flow right.
7. Schedules, plans, productivity and utility templates, etc.
8. … and plenty more.

I also don’t hold back on office purchases (quality desk, keyboard, chair, etc).

This ensures I have a friction-free experience.

It’s not much of an expense, but I see so many programmers on Quora trying to save a few pennies here and there and don’t realize it’s costing them far more in the convenience they lose out on.

I also don’t hold back on cloud computing.

My workflow is so much smoother when I am able to take advantage of much faster machines with more resources.

Here’s what I pay:

• $0.29/hr gets me instant access to a 32-core CPU (preemptible).
• For a uni assignment, this meant I could conduct over 100 experiments per hour instead of just 3. This gave me much faster iteration speeds that allowed me to evaluate a lot more alternate options in a short space of time.
• $3/hr gets me 64 CPU cores with an NVidia T4 GPU.
• I can also get a V100 with an 8-core CPU for $3/hr if I want.

You might be thinking, “but I could just buy my own 32-core computer with an RTX GPU”, and you’d be right. But these computers can also access the internet at 1000 Mbps. Plus I can launch an instance with eight V100s if I want.

Until recently I’ve just been using the ssh features in my JetBrains IDEs (or Jupyter Labs for ML with Python), but last week they released Projector, which is sort of like an IDE equivalent of ssh or Emacs’ Tramp.

It’s amazing.

With Projector I get to run my entire IDE in the cloud, which improves the feel a great deal. The IDE still can’t run notebooks properly, but I can happily use Jupyter Labs for now, and PyPlot works fine in the IDE (though I’ve not tried animations).

Compilation speeds are so much faster. They already were fast over ssh, but since the IDE is operating from the server, I don’t have to have anything on my local machine at all or need to wait for my IDE to transfer files to the server anymore

And again, with my server accessing the internet at 1000 Mbps, it’s a doddle to download a library or tool whenever I need to.

Everything is just so damn blissful, and I am able to make decisions that just aren’t viable with a normal internet connection.

It also means I can carry this power around with even the shittiest laptops now. In fact, I can do this all from my tablet and Bluetooth keyboard while connecting to Projector via my Vodafone mobile broadband dongle!