David Imel, reporting for The Verge:

Google’s new HiLight notification LED on the Pixel 11 Pro is nearly useless. Out of the box, the only two things it can glow for are when the phone is face down and you’re interacting with Gemini, or when you get a call from a favorite contact. And even then, it can only glow one of five colors!

This is a massive wasted opportunity for a feature that used to be so beloved on older Android phones, so developers have wasted no time doing what Google won’t.

X user Dhananjay_Tech developed a new app called HiLight Studio, which lets you do practically anything you want with the HiLight LED array. Did you receive a new Slack notification that mentions a certain word? Make HiLight pulse in a rainbow array. Get a new text? Make HiLight spin in a wave of green. You can even control things like timing per cycle, brightness, intensity, and saturation.

I’ve neglected writing about the Google Pixels this week because I haven’t got much to say about them. Just like almost every other technology product on the market in the mid-2020s, they add some new generative artificial intelligence features to justify nudging the price up by about $100 amid a crippling memory shortage caused by those very artificial intelligence features. The new HiLight LED is about the most canonical example of this phenomenon. Google clearly wanted to add a flashy Gemini feature to this year’s flagship, so it added an LED to the back of the smartphone and gave it a clever name. Thank goodness for this developer for making something out of this.

The most heavily marketed AI feature on this line of Pixel phones is called Proactive Assistance. Here’s an excerpt from Imel’s fantastic review:

Perhaps the largest push in this direction is the Pixel 11’s new “Proactive Assistance” feature. This is a set of dynamic chips that appear when certain interactions happen on your phone, like a friend asking what your schedule looks like next week or if you booked that table at a restaurant yet.

For the most basic interactions, like checking availability or adding an event to my calendar, I found Proactive Assistance really helpful. I practically live inside my calendar, and text conversations about scheduling are common. When my friend Marc texted me asking when I was available next week, Gemini popped up and suggested checking my calendar for times. Outside of simple scenarios like that, though, Proactive Assistance gets less reliable. One of Google’s key examples is letting Gemini book a dinner reservation for you if you’re chatting about heading out to eat, but I had to receive a text with an extremely specific string of information to get it to work — and even then it only worked sometimes…

One promise of large language models is the ability to understand what you mean and remove the requirements for smart keywords, so it’s odd that explicit keywords are still necessary to get so much of Proactive Assistance to work. I can see what Google is trying to do with this and I can maybe see a future where it gets refined, but as it stands, it’s simply too inconsistent. Plus, I had to manually toggle on Proactive Assistance and add OpenTable to connected apps in settings, which I just don’t think most people will know to do. As my colleague Cameron Faulkner puts it, “You need to be proactive to get Proactive Assistance to work.”

Proactive Assistance sounds like a generative AI-powered version of Siri Suggestions, which, alas, haven’t been updated with Siri AI yet. Because they rely on Apple’s old machine learning frameworks, I’ve always found them to be almost useless. For years, I’ve crowned the Google Pixel the best Android phone on the market purely because of Google’s excellent Android skin, which some incorrectly call “stock” Android. (It isn’t “stock” — it’s Google’s well-done take on the Android operating system.) But some years, Google misses the mark with its software, and this seems like one of them. I’d expect a lot more from the company that has, for years, championed its smartphone as “the most intelligent.” Continuing from Imel’s review:

As is tradition, Google also introduced a bunch of new camera features this year, and perhaps the most prominent is Magic Capture. The feature is front and center in the camera app right next to the photo and video modes, which tells you something about how confident Google is in it. Start a Magic Capture, point it at what’s happening, and when you’re done recording, Gemini will do some processing and pull out what it considers the best photos and video clips from the session.

This is one of the least objectionable generative AI-powered photography features Google has added to its smartphones in a while — consider me pleasantly surprised. I’m always torn between taking a ton of photos of important events — which I know I’ll look at later — and capturing one long video, which feels more true to form, albeit harder to skim. Magic Capture allows people to do both, all automatically, and I think it’s quite delightful. I was initially anxious about the “processing,” but it really seems quite minor, just enough to eliminate motion blur when it really must be avoided to preserve the shot. I feel like Google — and other technology companies embracing generative AI — would enjoy much more success in the public eye by building features like this, not ones that aim to redefine photography altogether.