One of the things that I am diving much deeper (pun intended) into lately is deep learning. And with deep learning, one of the things that can help you the most when it comes to having the right hardware locally is a beefy Graphics Processing Unit (GPU). As it turns out, offloading tasks that involve matrix multiplication to the GPU yield massive performance benefits compared to doing the same thing on the Central Processing Unit (CPU).

Now that I got authentication out of the way, it's time that we actually get something useful done with the Halo Infinite API. This whole saga started with me wanting to get the match stats so that I can analyze them outside the game, and that's what I thought I'd tackle first.

A week ago I was finally able to figure out what endpoints the Halo Infinite Web API uses. Now, the challenge became figuring out how to properly request the data from those, as there were two component pieces to every request - a Spartan token, and a clearance. After fiddling with the API a bit, and looking at the endpoint that aggregates all other endpoints, I was able to learn that there is a straightforward way to get all the right tokens through a number of chained requests, that are documented in this blog post.

As with most of my reverse engineering stories, this one starts with "Hmm... I wonder if I can get this data anyway?" I mentioned this in my previous blog post that I just finished the Halo Infinite campaign, and the next step was multiplayer, which also meant that I wanted to keep track of my stats to see just how bad I am playing against real people and aimbots.

What better way to spend the couple of remaining vacation days than by diving into the new Halo campaign. Ever since I got an Xbox, Halo was a staple of my gaming collection. The first Halo I ever played was Halo 3: ODST, and it completely changed my perception of first-person video games.

I recently moved most of my websites over to Netlify, because, well - I work there, and I want to be dogfooding as much of our product as possible. As part of this, I enabled my sites to use Netlify Analytics, which has been a fantastic lens to look at the site usage from a server-side perspective.

Spotify's foray into podcasting may be fairly recent, but I'm already discovering some interesting APIs that I can play with. The podcaster dashboard is tremendously useful and offers way more data than Apple and Google combined (with better reliability too), so the more I use it, the more I started thinking that it would be helpful for me to build some kind of automation mechanism to ingest the data into my own storage and then process it outside the default dashboard boundaries. In the process, I also spotted a little thing that I wanted to share with readers, and that is the ability to enable experimental features inside your podcast private view.

Twitter just announced that they are re-launching their verification program, and now you can check whether you are eligible directly from your Twitter account settings. Neat! Which naturally made me curious as to what they used behind the scenes, since in previous instances of the process they used a form that likely was manually reviewed. Before I dive in, I want to call out - no, I am not important enough to be verified, nor do I care about getting the blue tick.

Not too long ago, I was reading Charlie Kindel's "You're Thinking of Your Career Trajectory Wrong" and it reminded me of yet another trope that somehow is very commonplace, at least in the tech industry - your career is not a sprint, it's a marathon. It comes from a well-intentioned desire to communicate the fact that careers should be looked at through the long-term lens, which makes sense - we rarely hear the story of someone being right out of college and then becoming a Vice President of Product at a Fortune 500. It's all about the long game. But it also sets a lot of folks for failure early on.

The recent saga with the botched remake of the Grand Theft Auto Trilogy reminded me of the topic of this blog post, and I figured - what better time than now to get it out of my brain and into an array of bits. The topic today is rented software. Writing this I really feel like "old man yelling at (software) cloud," but maybe by some miracle it'll grab the attention of those making these kinds of decisions.