TL;DR: Meta is reportedly planning its biggest layoffs since the 2022–2023 restructuring. Unlike last time, which was framed as a correction for pandemic overhiring, this round has everything to do with AI. If it happens, it makes Meta the latest in a slew of companies laying off workers while ramping up AI operations—and signals the efficiency push never really ended; AI just gave it a new cover.

What happened: Reuters reported on Friday that Meta is considering laying off around 20% of its nearly 79,000-person workforce—that’s a potential 15,000 jobs or more. It cited three anonymous sources, two of whom said top execs have already begun signaling the plans to senior leaders, asking them to start figuring out how to cut back. If the 20% figure holds, it would be the single largest round of cuts in the company's history—deeper as a share of headcount than either of the two rounds in 2022 and 2023, which totaled around 21,000 jobs across both. A Meta spokesperson called the report "speculative reporting about theoretical approaches," which is the corporate equivalent of not a denial.

Two forces, one outcome: There are two distinct drivers. The first is cost: Meta plans to nearly double its capex this year to as much as $135 billion, with the majority going to AI infrastructure—including a commitment to $600 billion on US AI efforts by 2028. The second is productivity: The cuts are apparently premised on AI tools enabling leaner teams to do more.

Meta’s AI play: The company has been spending aggressively on the other side of the ledger: acquiring agentic AI startup Manus for roughly $2 billion, buying the AI agent social platform Moltbook, and bringing in Alexandr Wang to lead its new Superintelligence Labs as part of a $14.3 billion investment in his company Scale AI—while reportedly offering some researchers pay packages of over $100 million to build out the team.

The company it keeps: This is part of a fast-moving pattern across Big Tech. Block cut 40% of its headcount in February, explicitly citing AI as the reason. Atlassian cut 10% of its workforce—about 1,600 people—earlier this month to fund AI investments and enterprise sales. Amazon eliminated around 16,000 roles in January, citing the need to reduce layers and bureaucracy as it ramps up AI spending.

Bottom line: Mark Zuckerberg called 2023 the "Year of Efficiency"—a name that implies a beginning and an end. What this round of potential cuts makes clear is that efficiency seemingly never ended. The new equation is: Cut labor to fund compute, then bet the compute pays for itself. —SM

Maximizing Claude with Connectors

Context is king. And when it comes to AI, that’s even more true. I’ve been using Claude for over a year now and have recently leaned into Connectors to give it more context. This makes Claude’s outputs more effective, comprehensive, and personalized.

Out of the box, Claude knows a lot about the world. But when you use Connectors to plug it into your other apps, it begins to understand your world: your meeting notes, Slack messages, and upcoming events.

How to do it: Open Claude → on the left pane, click Customize → then click Connect your tools → search the available tools and click the + button on each to add them. You can manage the settings of individual Connectors (and toggle them on and off) on this screen as well.

One of my favorite Connectors is Granola. Since I use this tool to take notes in almost all of my meetings, Granola is extremely rich in its understanding of my relationships with specific coworkers, projects I’m working on, and even my communication style. If I’m asking Claude to help with a project, it can use all of this context to make better recommendations and give better insights.

Connectors can also take action on your behalf, so it’s possible to keep a workflow like “Ask Granola about any follow-up meetings I was supposed to schedule for this week and then schedule them in my calendar” in Claude. As you plug in more Connectors, you gain context and streamline switching between multiple apps repeatedly. —KH (Looking for more AI tips? Check out this recent episode of my podcast, Per My Last Email.)

If you have a tech tip or life hack you just can’t live without, fill out this form and you may see it featured in a future edition.

Most people say they prefer human writing to AI’s. Last week, a New York Times blind quiz hooked 86,000 of us (and counting) up to a polygraph… and determined that’s (sort of) a lie. Readers picked between five sets of samples written by famous human authors and Anthropic's Claude. 54% chose the robot—a slim margin, but enough to get the internet debating if human scribes are cooked.

Both AI researcher Ethan Mollick and Reddit co-founder Alexis Ohanian seemed to treat it as a watershed moment. Writers, predictably, took it well. One journalist announced it was “time to walk into the ocean.” Others argued the results said more about the average person's reading ability (not great) and their taste (even worse) than AI's chops. And plenty questioned the quiz's sparse samples: context-free snippets that reveal little about whether they hold up across a page, let alone an entire book. (Novelist Lincoln Michel likened it to “comparing Hallmark card slogans to random lines of epic poems.”) And only one of the human samples was from the 21st century.

What the 54% does show is that we may not be as good at quickly telling humans from machines as we’d hoped. After you finish, the quiz tells you that AI writing is “much more fluid” and much less prone to “grammatical errors or nonstandard syntax” than before—now it’s human prose that contains “clunky phrases.” (The example NYT uses? Pulitzer Prize winner Cormac McCarthy.)

Still, if you're a writer worried about the market demand for human-penned text, hold off on pivoting to a career as a plumber—anti-AI bias may work in your favor. Studies have found that when people are shown a set of AI-generated artwork, but only one is labeled as AI, they tend to give that work a lower rating. Turns out the four most powerful words in the English language might be “written by a human.” After all, no one reads Blood Meridian because all the commas are in the right place. —WK

Chaos Brewing Meter: /5

• Finally, a new generation of AirPods Max is coming. They have a ton of new features and will come in five colors.
• No GMOs, no pesticides, no AI. Companies are reportedly racing to create a universal “AI-free” logo.
• Not content with just owning the GPU market, Nvidia is reportedly going to focus more on CPUs—and could announce details about its new agentic-optimized CPUs in the next few days.
• The next career being targeted for LLM training? Improv acting, for its “ability to recognize, express, and shift between emotions.”
• OpenAI’s advisors are apparently freaking out about it wanting to add X-rated chats, with one saying it risks creating a “sexy suicide coach.”
• How Pokémon Go could lead to the next wave of delivery robots.
• A study tested 81 headphones—every pair had traces of harmful hormone-disrupting chemicals.
• Could your job be replaced by AI? Search this chart at your own risk.