The Box Agent: unleash your enterprise content with AI

Good morning. It's Tuesday, April 14, and we're covering China’s rise, workplace AI adoption, OpenAI doubling down on its London expansion, and more.

OpenAI Expands London Office After Pivot
OpenAI is doubling down on its presence in the U.K. by opening its first permanent London office, even as its ambitious local AI infrastructure plans face setbacks. The new office, expected to open in 2027, will significantly expand the company’s footprint and could house hundreds of employees as it builds what may become its largest research hub outside the United States. The move comes shortly after OpenAI paused its proposed U.K. “Stargate” data-center project, citing high energy costs and regulatory challenges.

TSMC Profits Surge on AI Demand
Taiwanese chip giant TSMC is expected to report its fourth consecutive quarter of record profits, fueled by relentless demand for AI infrastructure. Analysts forecast the world’s largest contract chipmaker will post about T$542.6 billion ($17.1 billion) in net profit for the first quarter of 2026, roughly a 50% year-over-year increase. The surge is driven largely by strong demand for cutting-edge 3-nanometer chips and advanced packaging used in AI systems, which currently exceeds the company’s production capacity.

Apple Tests Multiple Smart Glasses Designs
Apple is reportedly experimenting with four different frame designs as it develops its long-rumored smart glasses. The prototypes include two rectangular styles and two oval or circular options, with potential color choices like black, ocean blue, and light brown. Unlike full augmented-reality headsets, the glasses are expected to be simpler, display-free wearables that focus on features such as taking photos and videos, answering calls, playing music, and interacting with Siri. 

Microsoft Experiments With OpenClaw-Style AI Agents
Microsoft is developing new AI agents for Microsoft 365 Copilot inspired by the open-source agent framework OpenClaw. These agents would be able to operate more autonomously, monitoring things like inboxes and calendars and proactively completing tasks without constant user prompts. The effort is aimed primarily at enterprise customers, with stronger security controls than the open-source version of OpenClaw, which has raised safety concerns.

New from Box: The Box Agent. A fundamental shift in how enterprises use content to autonomously execute complex tasks. Using the latest advanced reasoning models, the Box Agent helps organizations put unstructured data to work to unlock critical insights from their content, automate tasks, accelerate decisions, and help teams work smarter — while keeping sensitive content protected every step of the way.

Your business lives in your content. Learn how the Box Agent unleashes it.

In an April 13, 2026 opinion essay, Sebastian Mallaby argues that U.S. export controls on advanced semiconductors have failed to slow China’s artificial intelligence progress. Chinese firms have bypassed restrictions by renting overseas computing power, stacking lower-end chips, and rapidly copying leading models through “distillation.” Mallaby contends the U.S. and China are now roughly level in AI capabilities, with China potentially ahead in industrial deployment.

He proposes shifting strategy toward a global AI safety agreement—modeled on nuclear nonproliferation—to manage shared risks rather than attempting to contain China’s advances. → Read the full article here. (Paywall)

A Gallup poll finds growing but uneven adoption of artificial intelligence in U.S. workplaces, with about 3 in 10 employees using it frequently and roughly half rarely or never using it. Among those with access, two-thirds report productivity gains, especially in management, healthcare, and tech roles.

However, skepticism persists: many workers cite ethical concerns, data privacy risks, or simply prefer existing workflows. Concern about job loss is also rising, with 18% of workers saying automation could eliminate their job within five years, up from 15% in 2025. → Read the full article here.

Rising electricity demand from AI data centers is driving new interest in carbon capture technology for natural gas power plants, with at least five U.S. projects under consideration involving companies like Google, ExxonMobil, Chevron, and Meta. Long viewed as too expensive, carbon capture is gaining momentum as tech giants seek to meet climate goals while securing reliable energy.

Executives say deep-pocketed “hyperscalers” could help scale the technology, though no U.S. gas plants currently use it. Analysts estimate up to $80 billion could be invested if projects move forward, but high upfront costs and reliance on federal tax credits remain major hurdles. → Read the full article here.

Claude Mythos Alarms UK Regulators: Anthropic claims its unreleased model can exploit zero-day flaws across major systems; UK authorities plan urgent talks as experts debate whether risks are overstated or credible.

NYT Reshuffles Leadership Around AI Push: The New York Times promotes Alex Hardiman and Hannah Yang to executive VPs, expands AI and data oversight, and begins a CTO search as it doubles down on subscriptions.

Harvey Launches End-to-End Legal AI Agents: Startup unveils agents that plan, research, and draft legal work across practice areas, producing cited memos and reports while lawyers retain final judgment and oversight.

Vercel Signals IPO as AI Drives Growth: CEO Guillermo Rauch says the company is “ready” to go public as ARR jumps to $340 million, fueled by AI-generated apps, with agents now creating 30% of deployments.

Meta Set to Overtake Google Ads: Emarketer projects Meta will hit $243.46 billion in 2026 ad revenue, surpassing Google’s $239.54 billion, driven by Advantage+ tools and faster growth rates.

Ask it at the end of any plan, email, strategy, or decision. It consistently surfaces blind spots you didn't think to ask about directly.

A few others in the same spirit:

Thanks for reading. See you next time!

— Matthew Berman, Nick Wentz & the Forward Future Team