HARDWARE

How Apple silicon keeps quietly piling up AI wins

In 2026, the Mac mini has turned into a surprising symbol of the next stage of the AI revolution. 

The powerful little box makes it easy to run a personal AI agent that can work on your projects 24/7 and do it in a secure, sandboxed environment with plenty of performance. And it's roughly the same price as a Chromebook.

Apple can thank OpenClaw for kicking off the phenomenon, but the overwhelming demand for Mac minis that has pushed backorders out by four to five months is hardly a claw-only phenomenon. Other personal AI agents, such as Perplexity's Personal Computer, Hermes Agent, and China's "raising lobsters" phenomenon, have all tapped into the Mac mini as the perfect machine to run your always-on AI agent in the safest way.

The fact that Siri, Apple's much-maligned personal assistant, has continued to dominate the narrative about its AI strategy has obscured the fact that the company's deep investments in Apple silicon chips to run machine learning and AI workloads have turned its Mac laptops and desktops into essential gear for the people and companies building the current AI revolution. 

Apple silicon launched on the iPhone in 2010, before arriving on the Mac in November 2020. The key breakthrough was putting the CPU, GPU, Neural Engine (for AI workloads), and memory onto a single chip. The result was devices that became much faster and more energy efficient, while also emerging as a strong platform for AI, software development, and local AI workloads.

"I think the foundations of Apple silicon have put us in a good spot," Doug Brooks, Apple's product manager for Mac hardware, told The Deep View in an exclusive interview. "The performance, the unified memory, the power efficiency, the targeted acceleration to accelerate the most critical parts of the AI compute workloads have put all of our systems … in a position to be able to tap into this. And we've seen that particularly in the agentic space with the insane demand for Mac minis and Mac Studios." 

Some of the hardware wins that put Apple in such an enviable position include:

• AI developers go all-in on Mac: In the heart of the AI ecosystem in the San Francisco Bay Area, whenever I walk into frontier labs, AI infrastructure providers, or startups building their slice of the AI ecosystem, 90% of the employees are running Macs. And that includes most of the developers, other than a few running Linux. 

• Frontier labs launch projects Mac-first: Because so many of the developers at the labs are running Macs, when they launch their most important new desktop apps they tend to launch them on Mac first, with Windows to follow later. Significant examples from the past six months include Claude Cowork, OpenAI's Codex, Perplexity Computer, and Google Gemini desktop app.  

• Mac mini becomes the agent machine: Again, the OpenClaw phenomenon kicked off the Mac mini craze, even though OpenClaw founder Peter Steinberger never advocated for it and was even a bit up in arms about it. It was a community-based movement that rapidly gained a life of its own from developers and AI enthusiasts. 

• MLX wins developer support: Apple launched MLX in 2023 as an open-source project that allowed developers to optimize their AI and machine learning software on Apple silicon. This is essentially Apple's version of Nvidia's CUDA. It's now critical for on-device processing and local inference, which saves money on tokens and provides better privacy and performance. This will be critical in the next stage of AI.

• The appeal of on-device AI grows: Because agents eat up so many tokens, which drives up the costs of AI, and the AI boom is spreading to more parts of the economy, such as regulated industries, the value of on-device processing looks poised to become one of the big stories of the second half of 2026 and beyond.

Of course, developers and AI enthusiasts are a small niche that Apple has already won over. However, only 16% of the people of the world regularly use AI today. The next wave of AI users is likely to be using phones more than computers, and privacy and performance will be table stakes. Since the same Apple silicon architecture that powers today's Macs is also what powers the iPhone, that device is going to be equally poised to run models and agents locally on-device and become an AI-forward hardware platform. We'll see how Apple's transformation of Siri plays out over the summer and with the launch of iOS 27 this fall. No doubt, that will continue to be the primary measuring stick that the public uses to measure Apple's progress in AI. However that goes, Apple's hardware wins in the AI ecosystem are only likely to accelerate as privacy, performance, and token cost continue to grow in urgency.

TOGETHER WITH IBM

Helping to solve AI code generation consistency problems

As teams rely more on AI for code generation, consistency becomes harder to maintain. When there are no shared rules, generated code reflects individual preferences—which can lead to mismatched patterns, uneven tests, and pull requests that take longer to review.

The concern is measurable: 55% of engineering leaders surveyed worry about losing shared understanding, and 39% worry about shipping with confidence as AI‑generated output increases.

Project‑level rules can set clear expectations. They define the guidelines, coding standards, and conventions a project follows, helping teams stay aligned as codebases grow.

See how to standardize AI code

POLICY

OpenAI wants stronger AI guardrails than Washington

One of AI's most prominent companies and the White House have different views on how the tech should be governed. 

Following the Trump Administration's executive order requesting a window for government review of powerful new models, OpenAI has released a policy document calling for more stringent oversight than the US government is proposing. 

"Effective governance requires visibility into how frontier capabilities are evolving … Building that understanding necessitates creating institutions capable of evaluating frontier AI, monitoring how capabilities evolve, and providing policymakers with reliable information," OpenAI said in the document. 

The policy paper differs from the executive order in two key areas: which agencies are responsible for administering these pre-deployment safety evaluations, and whether they must be completed at all. 

• While the executive order calls for the National Security Agency, a defense agency, to govern these checks, OpenAI suggests that they be run by civilian agencies: The Center for AI Standards and Innovation, or CAISI, working under the Commerce Department’s National Institute of Standards and Technology, or NIST. 

• Additionally, OpenAI's paper calls for these checks to be mandatory, while the executive order poses these evaluations as a voluntary framework. 

Still, while OpenAI seeks to nudge the US government toward its regulatory framework, CEO Sam Altman said in a post on X that the executive order "gets the balance right" on the US leading in AI development while ensuring safety. 

Altman isn't the only one supporting the executive order. Rival Anthropic said in a post on X that the order is an important step in "strengthening America’s leadership in AI." IBM CEO Arvind Krishna said at the Axios AI+NY Summit on Wednesday that he supports the government's light-touch guardrails. Meanwhile, Stephen Schmidt, Amazon's Senior VP and Chief Security Officer, said in a post on Linkedin that the company welcomes the executive order. 

It makes sense that major AI companies are largely showing their support for this declawed executive order: For one, AI is largely losing the narrative among the American public, and these companies backing an executive order that subjects the tech to regulation (even though it isn't mandatory) could signal to the public that they care about the broader societal impacts. OpenAI suggesting that these checks be required is essentially taking this to the next level. Given that it operates the most popular consumer-facing AI product on the market, OpenAI pushing for stricter regulation may be an attempt at winning the court of public opinion. Rather than looking like the lab seeking growth at all costs, this proposal gave OpenAI the opportunity to appear as though it wants to go above and beyond its peers in safety.  

TOGETHER WITH AWS

AWS Summit NYC Has a Track for Every Builder

AWS Summit NYC brings together cloud leaders, AI practitioners, and AWS experts for a full day of sessions, workshops, and live demos.

Whether you're exploring agentic AI, infrastructure, or application modernization, there's a track built for you.

Register here for your free slot!

BIG TECH

How Google wants to turn prompts into companies

Behind the current AI boom is a developer community that both uses and builds powerful tools. Google recently released a wave of new developer-focused AI tools at Google I/O, but one person has gone further: fostering an entire online community.

Google's Logan Kilpatrick joined The Deep View Conversations podcast straight from Google I/O, not only to share more insights on the new products but also to draw on his rich quilt of experiences to address the broader AI space, the future of software engineering, the developer community and more. 

While Logan's official role is as a member of technical staff at Google DeepMind, he has become a well-respected voice in the developer community, accumulating over 320K followers on X and constantly engaging with users. Prior to joining Google, Logan worked for NASA, OpenAI, Apple, and other AI startups, making him a builder at his core.  

Topics covered in this episode:

• 'Vibe coding' compared to agentic engineering

• The capabilities of Google's new Gemini 3.5 Flash

• The internal "flywheel" at Google, where teams use AI to accelerate the development of products

• The differences between AI Studio and Project Antigravity

• The need for developers to regularly "reset their level of ambition”

If you want to understand the distinction between rapid prototyping and managing million-line production codebases with AI, this conversation will leave you much more knowledgeable about the principles of agentic engineering. 

📺 Watch on YouTube
🎧 Listen in your favorite podcast player

Subscribe to Deep View Conversations for interviews with the leaders shaping the future of AI, business, and technology.

LINKS

• Google tests control to help website owners show up in AI search results 

• EU unveils sovereignty package aimed at reducing reliance on foreign tech

• Google announces data center water stewardship commitments 

• AI music platform Suno raises $400M Series D at $5.4B valuation

• Anthropic unveils Services Track for Claude enterprise partnership network

• Qwen 3.7 Plus: Alibaba’s new multimodal agent model  

• Perplexity: It now connects Apple Health on iPhone

• NotebookLM: Create briefing docs, study guides, and blog posts on-the-go

• Android Fake Call Detection: Alerts users if it senses an impersonator

• Deloitte: Lead AI and Data Science Engineer - Manager

• AT&T: Lead Cybersecurity, AI Models, Frameworks & Implementation

• Paramount: AI Platforms and Orchestration Lead 

• PWC: NGA AI Engineer Senior Manager

GAMES

A QUICK POLL BEFORE YOU GO