Zuckerberg admits AI agent delays; Anthropic bets on Samsung chips and drug discovery
Meta's AI ambitions hit turbulence, Anthropic goes chip-shopping with Samsung, and Jersey Mike's reminds us the hype bubble is very real.
Subscribe free →Zuckerberg Tells Staff: The AI Agents Aren't Coming Fast Enough
At an internal all-hands, Mark Zuckerberg reportedly told Meta employees that the company's AI agent development has fallen behind his expectations. No specifics leaked on where the delays are, but this is notable given how loudly and publicly Meta has been betting on agents as the next platform shift.
Why it matters: When the CEO who's been spending billions on AI tells staff progress is slower than hoped, it's a rare moment of internal honesty that should temper the relentless hype narrative.
Anthropic Is Shopping for a Custom Chip — and Samsung Is at the Counter
Anthropic is in discussions with Samsung to develop a custom AI chip, following OpenAI's recent Broadcom chip partnership announcement. Nothing is signed yet, but the talks signal that Anthropic wants to reduce its dependence on Nvidia and control more of its own inference and training stack.
Why it matters: Every major AI lab is now racing to build custom silicon — whoever gets this right first gets a durable cost and performance edge that's very hard for competitors to replicate.
Anthropic Also Wants to Discover Drugs
At its "AI for Science" event, Anthropic unveiled Claude Science — a workbench for researchers that consolidates fragmented tools and datasets, generates figures, and is explicitly aimed at accelerating drug development. Anthropic framed it as AI's shot at dramatically compressing the timelines for scientific discovery.
Why it matters: Anthropic is pivoting from "safe chatbot company" to full-stack science platform, and if Claude Science delivers even partial results in pharma, the TAM expansion is enormous.
Jersey Mike's Filed an IPO Doc and It Mentioned AI. Of Course It Did.
TechCrunch went digging in Jersey Mike's IPO filing — a sandwich chain — and found AI mentions, because apparently no company can go public in 2026 without genuflecting to the algorithm gods. The piece uses it as a case study in how reflexive and meaningless AI disclosure has become in financial filings.
Why it matters: When a sub shop has to mention AI to satisfy investors, the buzzword bubble has officially left the building — and it raises real questions about how regulators and analysts should parse AI claims in public filings.
Google and Amazon's Net-Zero Pledges Are Getting Torched by AI Data Centers
A new analysis shows that AI's energy appetite is making it nearly impossible for Amazon and Google to hit their sustainability targets. Both companies had ambitious net-zero commitments, but the explosive demand from AI training and inference is blowing past their carbon accounting plans.
Why it matters: The environmental cost of AI is no longer an abstract concern — it's actively undermining corporate climate commitments at two of the world's largest tech companies, and regulators are paying attention.
Midjourney's Medical Scanner Still Has More Vibes Than Validation
Midjourney dropped a behind-the-scenes video of its dunk-tank ultrasound scanner — the mysterious medical device the image-gen company announced it's building for deployment in physical spaces. The video shows controlled phantom scans but conspicuously avoids any clinical validation data or peer-reviewed evidence that the thing actually works.
Why it matters: An AI company pivoting from generating fantasy art to medical imaging hardware is a wild enough move — doing it while withholding clinical proof of efficacy should raise serious flags for anyone tracking AI's expansion into healthcare.
Quick Hits
- →Meta quietly launched **Pocket**, an experimental app where users generate playable mini-games via text prompts — vibe coding meets mobile gaming. TechCrunch
- →Thiel Capital's Jack Selby has been snagging stakes in hot AI/chip startups like Etched through Arizona-based connections; his firm Copper Sky Capital is raising a $300M second fund. TechCrunch
- →World Cup host cities are deploying drones and mass surveillance cameras at scale — a real-world stress test for AI-powered public monitoring infrastructure. The Verge
- →New macOS malware **PamStealer** is making the rounds with unusually stealthy tradecraft, underscoring a growing trend of sophisticated infostealers targeting Mac users. Ars Technica
AI TLDR