Pixel 11 Pro Fold is the AI builder’s phone to beat

August 21, 2026

Welcome back. OpenAI’s new plugin for Apple Messages shows how quickly AI is shifting from answering questions to managing tasks, and it puts pressure on Apple's Siri overhaul coming in September. Meanwhile, AI’s limitless appetite for compute is running into constraints around energy, infrastructure, and resources, shifting the industry's focus to efficiency. And after a week with Google’s Pixel 11 Pro Fold, I’m convinced its combination of Wispr Flow-like voice dictation and a tablet-sized screen for remotely steering agents makes it the best phone for AI developers and enthusiasts right now. —Jason Hiner

CONSUMER

Two things make Google's Pixel 11 Pro Fold the best phone for AI builders. And it's likely to beat anything Apple launches in 2026.

I've been using Google's latest Pixel Fold device for the past week, and it's surprised me. I've tried a lot of foldables, reviewed them, and traditionally slid them back in a drawer when I was done. But the new Pixel Fold is going to be very difficult to put down, for two reasons:

Rambler voice dictation: The Pixel 11 Pro Fold has Google's new AI voice technology built into the Google Keyboard (Gboard). You simply tap the new mic + Gemini button and start rattling off your AI prompt (or any other text), and Rambler will remove the ums and ahs, adjust corrections you make midstream, and produce fully coherent, neatly formatted text. This is very similar to the Wispr Flow app on Mac (and Windows), which has been the AI industry's favorite productivity boost of 2026. Rambler might be even faster and more accurate than Wispr Flow on Mac. While Wispr Flow has Android and iPhone apps, they're clunky and cumbersome. Rambler works like a champ and is far better than the traditional dictation features built into Android or iPhone. For daily Wispr Flow users, it's a massive relief to be able to dictate on a phone like we can on a computer. Rambler is also on Google's other Pixel 11 devices, and could come to other Pixel and Samsung devices in time.
Controlling agents from the larger mobile screen: One of the things you can do once you're able to rattle off long prompts or instructions from Rambler is to open your AI apps and start new queries and tasks, or steer ones that are already in progress. With the interior 8-inch display that feels like a tablet, you can work in a mode that's much closer to a laptop or desktop. With ChatGPT, Claude, Gemini, and Perplexity, you can open the app on the inner 8-inch screen and get your list of chats and left menu-bar items on one side of the screen, with a full chat window on the other. On all four of those apps, you can also connect them to the desktop version of the app and remotely control the tasks your agents are carrying out. This is a great thing for AI developers at a time when there are reports of people walking around San Francisco and New York with their laptops open so their agents can keep running. The combination of agents being remotely controlled from your phones and the latest foldables offering an even more desktop-like experience from mobile could let developers touch grass a little more.

While I've been testing the Pixel 11 Pro Fold, I've also been testing Samsung's new passport-sized Fold 8, which is based on the long-rumored form factor of Apple's forthcoming iPhone Ultra, expected to arrive in September. The passport form factor is charming, very small and lightweight, and great for typing on both the cover screen and the inner screen. It's much smaller than the Pixel 11 Pro Fold. It's akin to the difference between the iPhone 17 Pro and the iPhone 17 Pro Max, but the size differential is greatly amplified on a larger screen. And then there's Rambler. The Fold 8 doesn't have it, although it might get it eventually. And Apple's Siri dictation is nowhere close to being as effective as Rambler or Wispr Flow. However, in January, Apple quietly acquired Israeli startup Q.ai, which could offer similar functionality. But until we hear otherwise, the size of the Pixel 11 Pro Fold screen plus its Rambler dictation feature put it a step ahead of Apple and Samsung for AI builders.

TOGETHER WITH JUMPCLOUD

Agents are onboarding themselves, accessing systems, executing workflows, and making decisions with no formal record, no named owner, and no offboarding when their purpose expires.

New research from JumpCloud found that non-human identities now outnumber humans in 83% of organizations, yet only 21% have governance controls in place.

The workforce has changed. The infrastructure managing it hasn't. HR for Agents isn't a metaphor, it's the missing operational layer for every organization running AI at scale.

See what securing every identity looks like .

HARDWARE

AI is already causing a resource crunch. But we may only be at the beginning, one executive says.

Kevin Cochrane, chief marketing officer of neocloud Vultr, told The Deep View that AI demand has only hit the "tip of the spear." Because most of AI demand is coming from hyperscalers, developers and AI-native startups, several massive companies are still very nascent in their AI journey.

For every Fortune 500 company, there is likely a portfolio of thousands of applications for both internal and external uses, according to Cochrane. "Every single one of those applications needs to get modernized," he said. "That's a 10-year build cycle, and that's going to demand a tremendous amount of compute capacity all around the globe."

It's no secret that enterprise could represent a cash cow for AI. That's why OpenAI and Anthropic are vying to serve major organizations, investing in forward-deployed engineers and signing partnerships with legacy tech organizations such as IBM to make that transition smoother.

And even beyond enterprises, Cochrane asserted, AI presents the opportunity for "unbounded demand," essentially only limited by humanity's capacity for creativity. "I think we're going to get to a point where we're going to see that there is a predictable pace of growth of demand," he said.

However, once that deployment comes to fruition, the industry runs into another problem: Compute capacity and energy constraints. There is a "theoretical limit" to how much power you can and should generate, or how much land and water you can and should use, Cochrane said.

With AI already impacting infrastructure, "something has to give," he said. The thing that gives may be the chips themselves, something that companies like Nvidia and AMD apparently realize as they shift focus to efficiency.

"Given the fact that there has to be some natural limit, and there's infinite demand … You have to be as efficient as possible," said Cochrane. "You have to drive down that efficiency curve year after year after year."

Cochrane isn't the only one reading the tea leaves on how AI will affect infrastructure. Investors have taken an interest in the technologies at the foundation of AI, such as energy and mining, seeking to get ahead of the problems that AI could create.

Cochrane is striking a common theme in saying AI will drive fundamental changes in work and life. Fundamental changes, however, require innovation from all sides. Chips and data center architecture are just one piece of the puzzle in making this technology more efficient. At the utility level, more sources of clean energy could ease the burden that AI and data centers are placing on overstressed energy grids. And the models themselves could get more efficient, too, such as Pathway's 150-million-parameter small reasoning model, BDH-CQ, which costs a fraction of OpenAI's most cost-effective model, GPT 5.6 Luna and demands less compute and energy. Efficiency is a core part of innovation. Computers used to take up entire rooms, and mobile phones used to be giant bricks as long as your forearm. AI's version of this means trimming down every component of the stack, from energy to infrastructure to the models themselves.

TOGETHER WITH TABS

Hybrid and usage-based models are changing how companies price, and how finance teams recognize revenue, forecast, and close the books.

In this on-demand session, Tabs co-founder Rebecca Schwartz and PwC partner Amit Dhir break down the real operational impact, with frameworks and examples you can apply now.

Watch the session

PRODUCTS

For over a decade, composing, searching, and sending Apple Messages with AI meant relying solely on Siri. OpenAI just changed that.

On Thursday, the AI lab launched an Apple Messages plugin that lets ChatGPT assist with your text messaging needs, including finding specific messages, summarizing conversations, and composing and sending messages. At launch, it is limited to ChatGPT on your Mac via ChatGPT Work and Codex.

While the description may make the plugin sound like just another feature that lazily helps you write or summarize a thread, the real-world help it can offer is far greater. For instance, in the demo video, a user asks ChatGPT which message threads from the previous day it should follow up on and gets a bulleted response that identifies each message and a suggested reply, which users can review and send. Another helpful use case is asking it to check your calendar and compose a text with a couple of free times.

The big question about this plugin's value is less about the help it can offer users and more about how it differentiates itself from the AI users will get with the launch of iOS 27. Although that won't happen until the iPhone launch event and the broader release of iOS, both likely in September, Apple previewed improved Siri features at its WWDC event in May, which The Deep View covered in depth. The updates include a chatbot-like interface for Siri that would let users ask it to do everything the OpenAI plugin promises.

Apple has yet to respond to a request for comment.

The major difference is that with the Apple version, users can use it across their ecosystem of devices, not just on their Mac, which is far more convenient. The only argument in favor of the plug-in is that ChatGPT power users won't want to context-switch between the two AI interfaces. While that is true, with the launch of the new and improved Siri, users can also just ask Siri the same questions conversationally and get the answers, circumventing the need for ChatGPT altogether.

The reaction to this plugin on social media has been overwhelmingly positive, with people calling it a game-changer and celebrating OpenAI for unveiling an AI feature that could actually help in everyday life. This has been really interesting to see, as I didn't see quite as overwhelming a reaction during WWDC, which could be because it was just one of the many features announced. Both this announcement and the positive reaction to the foldables launched this summer by Samsung and Google are teeing up Apple to unveil its own counterparts in September. The momentum also confirms one thing: this iPhone launch event may be one of Apple's biggest in years. It's one to watch for both new AI developments and to see which new AI features Apple puts in the spotlight.