🛎️ Enterprises Are Livid

Good Morning, AI Enthusiasts!

The models are expensive, the agents are late, the data centers need neighbors, and the café still wants its olive oil back.

TOKEN

Palantir CEO Says Enterprises Are Livid

👀 What's happening: Palantir CEO Alex Karp went on CNBC and said AI labs have “completely, irresponsibly, oversold” their models. His complaint was blunt: enterprises are paying for tokens, getting little value, and watching labs absorb their IP and market “alpha.” He framed himself as the public mouthpiece for CEOs who are furious in private but too cautious to say it on camera.

🌍 How this hits reality: Karp is selling Palantir, obviously, but the attack lands because it hits the ugliest part of enterprise AI. The pitch was automation. The invoice says tokens. The risk clause says your knowledge may become training exhaust. That is why “AI sovereignty” suddenly sounds less like nationalism and more like procurement self-defense. Companies do not just fear bad models. They fear paying vendors to commoditize the thing that made them worth buying from.

🛎️ Key takeaway: The AI labs sold intelligence as a utility, but enterprises are discovering the meter runs on their money, their data, and possibly their moat.

AGENTS

Meta’s Agent Hangover

👀 What's happening: Meta CEO Mark Zuckerberg told employees that AI agents are moving slower than expected. After cutting about 10% of the workforce and shifting roughly 7,000 people into AI teams in May, he admitted the restructuring was not “clean” and that its bets “haven’t come to fruition yet.” The new deadline is familiar: wait another three to six months.

🌍 How this hits reality: Meta did what Big Tech increasingly wants to do: spend like an infrastructure company, manage like a panic room, and ask software to justify human displacement after the fact. The company may spend up to $145 billion on AI infrastructure this year, yet its agents still cannot deliver the productivity story fast enough. The mouse tracking scandal makes the irony nastier. When automation fails to prove itself, the workforce becomes the dataset.

🛎️ Key takeaway: The agent boom’s awkward truth is that giants are reorganizing around an efficiency curve before the tools have earned the chart.

NEW TECH

Nvidia Kills AI Cooling

👀 What's happening: Nvidia announced a closed-loop liquid cooling system for its Rubin chips that runs coolant at 113°F, hotter than a hot tub, and exits at 131°F. Because that base temperature sits above most ambient air, data centers can dump heat with outdoor dry coolers instead of chillers, which eat nearly 40% of a facility's power. Nvidia claims up to 100% less water use: filled once, sealed for the building's life. Earlier this year, local governments delayed over 75 data centers over water and grid strain.

🌍 How this hits reality: This is a real dent in AI's two ugliest costs. Water use drops toward zero, and retiring the chillers frees up a 40% slice of power, with every warmer degree Vertiv, a data-center infrastructure supplier, counts saving another 4%. That directly answers the drained-aquifer and strained-grid complaints that stalled 75 builds this year. The single hard power problem left is the compute itself.

🛎️ Key takeaway: Nvidia just made the data center a far better neighbor, and that matters more than the engineering for now. The GPUs still run hot on the grid, but everything wrapped around them finally sips instead of gulps.

AGENTS

Two AIs Ran a Café Into the Ground

👀 What's happening: Andon Labs handed a real Stockholm café to an AI manager named Mona, running Gemini 3.1 Pro, then swapped to GPT-5.5 mid-June. Mona controlled pricing, purchasing, menu, and staffing, wired straight to the till. It approved a customer's fake 99% discount instantly, comped anyone who asked, and stockpiled 15 liters of olive oil for a shop with no stove. Two months in, the account fell from $40,000 to $10,000. GPT-5.5 only reversed the failure: it froze purchasing and let a quarter of the menu go out of stock.

🌍 How this hits reality: GPT-5.5 is a weak seat here; frontier models like Fable 5 and Opus 4.6 reason far better and would not fumble basic budgeting the same way. And Mona itself is a bare harness, primitive next to mature agent frameworks like OpenClaw or Hermes that enforce spend limits, route big decisions to a human, and refuse a 99% discount before the model ever sees it. Pair a stronger brain with a real framework and the café likely survives. Neither piece was present here.

🛎️ Key takeaway: The race sells smarter brains, but a business runs on the framework around the brain. This café didn't lack intelligence; it lacked everything built to keep intelligence accountable.

TOGETHER WITH MYCLAW

Run Best-in-Class AI Agents — In Seconds, Always On

MyClaw is a best-in-class AI agent platform that brings together OpenClaw, Hermes Agent, and commercial infrastructure to help businesses launch intelligent agents that understand goals, use tools, automate workflows, and deliver finished results.

We already host OpenClaw and Hermes — and we're just getting started. Claude Code, Codex, and every agent worth running are coming next. One home for all of them, always on.

Stop babysitting infrastructure. Spin up your agent, walk away, and let it work around the clock.

MyClaw — Your Agents, Hosted Right.

DAILY TL;DR

• Microsoft’s leaked Aion concept shows a lightweight Copilot-first OS built around agents instead of apps.
• Anthropic is discussing a custom AI chip with Samsung, signaling it may want more control over the inference stack.
• OpenAI proposed donating 5% of its equity to a U.S. sovereign wealth fund, pulling frontier AI deeper into state-capital politics.
• Microsoft launched a $2.5B AI deployment company, turning enterprise AI adoption into a services business.
• Meta quietly launched Pocket, a vibe-coded gaming app testing whether prompt-made games can become a consumer product.
• Nvidia outlined a capital-partner model to finance AI compute, making infrastructure funding part of its growth machine.
• Google is making original recipe links more prominent in AI Mode, another sign AI search is being forced to answer publisher pressure.

READ MORE

Let the Future Come to Your Inbox

Stay ahead without drowning in information. We turn the most important signals across AI, tech, marketing, and future products into 5-minute reads you can actually finish.

• AI Secret uncovers what really matters in AI
• Bay Area Letters decodes tech and business shifts from Silicon Valley
• Robotics Herald tracks how robots move from labs into daily life
• Marketing Secret breaks down real growth and go-to-market playbooks
• The Hardwire explores hardware, consumer tech, and what’s coming next

TOGETHER WITH US

AI Secret Media Group is the world’s #1 AI & Tech Newsletter Group, reaching over 2 million leaders across the global innovation ecosystem, from OpenAI, Anthropic, Google, and Microsoft to top AI labs, VCs, and fast-growing startups.

We've helped promote over 500 Tech Brands. Will yours be the next?

Email our co-founder Mark directly at mark@aisecret.us if the button fails.