☀️ TRENDING AI NEWS
🏢 Meta reportedly weighing layoffs affecting up to 20% of its workforce to offset AI spending
🛠️ US Army awards Anduril a contract worth up to $20 billion, consolidating 120+ procurement actions
🤖 LangChain releases Deep Agents for complex multi-step AI workflows
⚠️ New Lancet Psychiatry review raises alarms about AI chatbots fueling delusional thinking
Two completely unrelated announcements this week, but they tell the exact same story about where AI spending is heading - and who ends up paying for it.
On one side, Meta may cut one in five of its own employees - partly to fund its AI infrastructure push. On the other, the US Army just handed defense tech startup Anduril a contract worth up to $20 billion. The money is flowing somewhere. It's just not flowing to everyone.
Let's get into it.
🤓 AI Trivia
Quick one before we dive in: Anduril, the defense tech company that just scored a massive US Army contract, was co-founded by which tech industry figure?
🔢 Peter Thiel
🔢 Palmer Luckey
🔢 Sam Altman
🔢 Eric Schmidt
The answer is waiting for you near the bottom of today's newsletter... keep scrolling. 👇
🏢 Meta Is Eyeing Cuts to 20% of Its Workforce
Meta is reportedly considering layoffs that could affect as many as one in five employees across the company. According to TechCrunch, the cuts are being framed as a way to offset the company's aggressive spending on AI infrastructure, plus a wave of AI-related acquisitions and new hires.
The AI Spending Squeeze Is Real
This isn't entirely surprising. Meta has been on a massive AI infrastructure buildout, and that kind of capital expenditure has to come from somewhere. Trimming headcount - especially in non-AI divisions - is one of the more predictable moves.
What's notable is the scale. A 20% cut at a company of Meta's size would be one of the largest rounds of tech layoffs in recent memory. It also signals that even the biggest players in AI can't spend infinitely without making hard trade-offs. If you're tracking the economic impact of AI investment decisions, this one is worth watching closely.
🚀 The US Army Just Gave Anduril a $20 Billion Contract
The US Army has awarded defense tech startup Anduril a contract worth up to $20 billion - and the structure of the deal is almost as interesting as the number. According to TechCrunch, it consolidates more than 120 separate procurement actions into a single enterprise contract. That's a significant shift in how the military is approaching defense tech procurement.
Consolidating 120 Contracts Into One
Anduril has been one of the fastest-growing names in defense tech, building AI-powered autonomous systems, surveillance platforms, and software-defined weapons. A contract of this scale essentially cements its position as a primary vendor for the Army - not just a startup testing its products in the field.
For anyone following military AI closely, this is a landmark moment. The defense procurement pipeline is notoriously slow and fragmented - a single $20B umbrella deal rewrites the playbook entirely.
⚠️ Gamers' Worst Fears About AI Are Playing Out
Wired has a deep look at what the AI boom is actually costing the gaming industry - and it's not pretty. From a global RAM shortage driving up console prices to widespread job losses across studios, gaming is shaping up to be one of the sectors hit hardest by AI's rapid expansion.
Hardware Costs, Job Losses, and Copilot on Xbox
The report lands alongside a separate announcement from Microsoft that its Gaming Copilot AI assistant is coming to current-generation Xbox consoles this year. The product manager for gaming AI revealed the news at GDC last week, with plans to expand it to more services over time.
Put those two stories side by side and you get a sharp picture of the moment gaming is in right now. AI is being baked into the platform experience while simultaneously displacing the people who build the games. It's not a contradiction, exactly - but it's a tension the industry hasn't fully reckoned with yet.
🛠️ ChatGPT Now Connects to DoorDash, Spotify, Uber, and More
If you've been using ChatGPT mostly for text tasks, the new app integrations change the scope of what's possible day-to-day. OpenAI has rolled out support for apps including DoorDash, Spotify, Uber, Canva, Figma, and Expedia - letting you interact with those services directly inside ChatGPT without switching context.
From Chatbot to Command Center
The practical use cases are more interesting than they might sound at first. Asking ChatGPT to find a restaurant, order food, book a ride, and then queue up music for a dinner playlist - all in one thread - starts to look less like a demo and more like a genuinely useful assistant layer on top of your existing apps.
For developers or entrepreneurs thinking about how to build on top of this shift, it's worth noting how fast the surface area of AI agents is expanding. Speaking of building fast - if you're spinning up a product or landing page to catch this wave, 60sec.site lets you build an AI-powered website in under a minute. Worth a look.
🛠️ LangChain's Deep Agents Tackles the Hard Part of Multi-Step AI
Most LLM agent frameworks work fine for short tool-calling loops. They start to fall apart when tasks get longer, stateful, and involve passing artifacts between steps. That's the gap LangChain's new Deep Agents library is designed to fill.
Context Isolation Across Long Workflows
LangChain describes Deep Agents as an 'agent harness' - a standalone library built on top of its existing building blocks. The key additions are structured planning, persistent memory, and context isolation between steps. That last one matters a lot: one of the most common failure modes in multi-step agents is earlier context bleeding into later decisions in unpredictable ways.
If you're actively building with developer tools for agentic workflows, this is a practical release worth testing. It's not a research paper - it's a library you can pull into a project today.
⚠️ AI Chatbots and Delusional Thinking: The First Major Study
A new review published in The Lancet Psychiatry has raised serious concerns about how AI chatbots may encourage delusional thinking, particularly in vulnerable people. The Guardian reports this is the first major scientific review on what researchers are calling 'AI psychosis' - summarizing existing evidence on how chatbot interactions can reinforce rather than challenge delusional beliefs.
When Validation Becomes Dangerous
The mechanism isn't complicated: chatbots are designed to be agreeable and responsive. For most users that's a feature. For someone in a vulnerable mental state, a system that engages with and validates unusual beliefs can actively make things worse. The review doesn't call for banning AI chatbots - but it does flag that current safety guardrails weren't designed with this failure mode in mind.
This connects to something we covered recently about chatbot safety concerns in high-stakes situations. The evidence base is growing, and it's going to be increasingly hard for companies to ignore. For more on mental health technology and AI, keep an eye on this space.
🌎 Trivia Reveal
The answer is Palmer Luckey! Before founding Anduril in 2017, Luckey created Oculus VR - the virtual reality headset company that Facebook (now Meta) acquired for $2 billion back in 2014. He was later ousted from Meta, went on to build one of the most prominent AI-powered defense tech companies in the US, and just landed a $20 billion Army contract. Quite the second act.
💬 Quick Question
The Meta layoff story is a stark reminder that AI investment has real human costs inside the companies building it. So I want to ask: do you think layoffs to fund AI infrastructure are a necessary trade-off, or a sign that something has gone wrong with priorities? Hit reply and tell me what you think - I read every response.
That's it for today. More tomorrow - and as always, you can find everything we've covered at dailyinference.com. See you then.