Happy Friday. I'm back from vacation and still getting caught up on everything I missed. AI researchers moving jobs is getting covered like NBA trades now, apparently. Before I get into this week's issue, I want to make sure you check out my interview with Perplexity CEO Aravind Srinivas on Decoder this week. It's a good deep dive on the main topic of today's newsletter. Keep reading for a scoop on Substack and more from this week in AI news.From chatbots to browsersSo far, when most people think of the modern AI boom, they think of a chatbot like ChatGPT. Now, it's becoming increasingly clear that the web browser is where the next p …Read the full story at The Verge.

I'm not ashamed to admit that Google's AI-powered personalized research assistant, NotebookLM, is the only AI tool I heavily rely on. One of the reasons the tool works so well for me is because all its features seamlessly fit into my workflow. I'm a full-time student, and a lot of NotebookLM's features seem like they've been made with students (especially students like me) in mind.

Between the bulk of the '80s and the '90s, DOS was the way to go for gamers, and was the backbone of PC gaming. Sure, there were plenty of other consoles and platforms with tons of games, yes, but DOS was where the most innovation was happening. We're talking about experimentation with 2.5D and 3D space, first-person shooters on PC, massive city-building sims, point-and-click adventures, and so much more. Devs were building games for DOS while pushing boundaries, experimenting with new perspectives, mechanics, and genres, effectively laying down the groundwork for modern gaming as we know it today.

When it comes to PC hardware and performance, you'll find countless conflicting opinions, depending on how many forums or videos you check out. From choosing the right components and optimizing them correctly to maximizing performance, it can often get tricky to have a north star of your own. This leads to numerous misconceptions and bad purchase decisions that are not always easy to come back from. Hence, I've compiled the most common PC performance misconceptions you need to be aware of.

For smart home deployments, there are a lot of ways to track the presence of individuals in a room. There are motion detection sensors that use mmWave and PIR, and you can also use simple detection mechanisms like a door opening or closing with a door sensor. However, to detect continued presence, things get a bit more complicated. Many sensors claim to be able to detect continued presence, but I've found that sitting still (common when lying in bed or watching a movie on the couch) can trick them easily. That's why I built a presence sensor with an ESP32, and it works perfectly.

How big a monitor is too big? Trick question; there's no such thing as a monitor that's 'too big'. However, the bigger the screens get, the higher the price rises, until you're looking at some pretty eye-watering MSRPs for some of the largest screens out there. So, when one of these screens goes on a huge discount, you're essentially getting all of the benefits of a huge monitor without the downside of a giant price tag.

Running Frigate on hardware is more straightforward than you may immediately believe. It's not a taxing package, but running IP cameras and using detection algorithms is where your CPU can really bog down. Thankfully, most recent and moderately powerful mini PCs are up to the job. But what if you're building a home-based data center?

They say that hindsight is 20/20, and I couldn't agree more. At the moment, I never felt I was overspending on any of these components. In fact, I remember it as making smart decisions, buying only what I "needed," and spending more on the parts that deserved it. It's only after a long time has passed that I consider these purchases more expensive than they needed to be. Some of it has to do with bad timing, but most of it was me being impulsive. So, here are the five components that I spent way too much on, and what I should have done instead.

It's almost human nature to want to make the most out of our devices at this point, and no matter what, we're always looking at how we can optimize our computers (or even phones). One way we tend to do that is by reducing memory usage, but it's not always obvious how you can do that.

A trend in gaming that has mostly gone away is the special peripherals, usually in the form of novelty controllers, that sometimes come out alongside new games. The most popular of these, and the one responsible for their decline, are the Guitar Hero guitar controllers, which are almost life-size and shaped like a guitar. There are two major components that contribute to the cult classic status of some of these, which are usability and novelty. The best controllers were often one or the other, but the cream of the crop managed both. These days, the closest we have are the Nintendo Switch 2 Joy-Cons, or some of the classic controllers available for the Switch, like the GameCube controller. For this list, I went with ones that managed to do both, at least for some games.

Every developer has their go-to tools, and for me, VS Code is the undisputed champion when it comes to Python. But it’s not just the editor itself that makes the difference; it’s the carefully curated collection of extensions that turns a good coding environment into a great one. I have experimented with many, and in this post, I’m eager to share the extensions that I use every single day to make coding in Python seamless, intuitive, and more fun.

The advent of camera phones spawned smartphones, instantly ballooning the storage requirements on handheld devices in the past couple of decades. Most users are now inexplicably dependent on cloud storage to manage documents, back up their app data, and even store copious amounts of images and video. This is no accident, and corporations happily mint money from cloud storage services without subsidizing hardware for consumers in return. In fact, I'm tempted to believe the death of expandable storage on smartphones also contributed to a boom in cloud storage demand.