🛎️ 900 Billion Agents Will Eat the Internet

Good Morning, AI Enthusiasts!

The industry may be building the first economy where the customer base can be spawned programmatically.

AGENTS

900 Billion Agents Will Eat the Internet

👀 What's happening: Huawei's Intelligent World 2035 report, Trend 9, promises 9 billion people and 900 billion AI agents by 2035. One hundred per human. Its own numbers erase the humans anyway: agents will move 90 percent of AI token traffic and drive a 100-fold rise in data volume. The same week, Reuters reported agents will move over 90 percent of AI token traffic by 2035. Hyperscalers had already committed $725 billion of capex for 2026, up 77 percent in a year.

🌍 How this hits reality: Read the order of operations. The money is already spent, and the depreciation clock is running. So the industry is not betting that people will want more AI. That demand is capped: a person reads five words a second, then sleeps. The only world where the capex closes is one where machines buy from machines, agents hiring agents, spawning agents that need compute only more compute can answer. That is the bet, and it has nobody in it.

🛎️ Key takeaway: This is where agents are going: not tools for people, but customers for machines. The industry is not betting on you. It is betting on your absence.

Full research: https://100-agents-per-human.showly.site/

TOGETHER WITH SHOWLY

The First Agent Artifact Cloud

Every publishing tool ever built assumed a person at the keyboard. Showly drops the assumption. Your agent researches, writes, designs, and publishes the page itself, as real HTML, at an address that exists the moment it is generated. No editor. No domain. No DNS. No deploy.

Codex, Claude Code, OpenClaw, Hermes, Pi, MyClaw: whatever your agent is, it gets somewhere to put the work.

Nothing between the idea and the showcase.

A free plan is available right now — your agent can have somewhere to publish in seconds.

PREVIEW

The Brake Is for Other People

👀 What's happening: Days ago, Anthropic CEO Dario Amodei published a long essay urging the industry to slow down. Sam Altman, Elon Musk and Demis Hassabis nodded along in public. Days later the leaks landed: OpenAI's GPT-6 Sol and Anthropic's own Claude Opus 5.2, both surfacing in test accounts. Testers report faster answers, cheaper tokens, models that check their own work. OpenAI's Greg Brockman teased a breakthrough in pure mathematics.

🌍 How this hits reality: How noble. Now check the calendar. The caution was published, and six days later its author was shipping a new frontier model. That is not a brake. It is a starting gun that clears its throat first. Every lab wants the others careful and itself fast. The fastest way to slow a rival is to hand him a reason to wait.

🛎️ Key takeaway: The letter was not written to slow the industry. It was written so the other guy slows first. Nobody did. In this race, caution is a lane change.

BILL

AI's First Guardrail Is a Utility Bill

👀 What's happening: The House passed the Ratepayer Protection Act on Wednesday, 417 to 3, the first major legislation on the costs of the AI build-out. Co-sponsored by Rep. Gabe Evans, a Colorado Republican, and Rep. Kathy Castor, a Florida Democrat, it asks states to consider a standard making data centers cover 100 percent of the new power plants and lines they require. It heads to the Senate three weeks before the midterms.

🌍 How this hits reality: Notice which argument moved Congress. Not AI safety, debated for years with nothing to show. Electricity bills. The bill's own mechanism is soft: states consider, but need not adopt, and Senator Martin Heinrich calls it basically voluntary. It still passed 417 to 3. When the complaint is a monthly bill, Washington answers in weeks. When it is risk, it answers in essays. The first real guardrail on AI is not ethical. It is financial.

🛎️ Key takeaway: Safety asks the public to accept a risk it cannot see. A utility bill arrives every month. That is why AI's only real rule is about money.

Full research: https://bending-spoons-research-en.showly.site/

NEOCLOUD

Neoclouds Are Now an Asset Class

👀 What's happening: Ten banks are lending $22 billion to Crux AI, the cloud venture of Blackstone and Alphabet, launched last week. Goldman Sachs, Sumitomo Mitsui, Barclays and BNP Paribas lead the group. The debt buys Google's TPU chips, secured by those chips and by Crux's customer contracts. Blackstone added $5 billion of equity. The lenders are syndicating the debt, and it could later be replaced by investment-grade bonds.

🌍 How this hits reality: Look at the size against the sector. Nebius, the biggest prior neocloud raise, took 5.75 billion. CoreWeave took 8.5 billion. Crux just took 22, from a standing start. That is not a company outgrowing its peers. That is a market repricing the whole sector in a single deal. Collateral, contracts, syndication, bonds: the structure is a template, and the template just tripled in value.

🛎️ Key takeaway: Nebius was a bet. CoreWeave was a bet. Crux is a market. Wall Street stopped lending to neoclouds and started pricing an asset class.

DAILY TL;DR

• Salesforce launched Koa, a CRM reasoning model built on Nvidia Nemotron and trained on synthetic enterprise workflows rather than customer data.
• Anthropic signed its first Australian data-center lease for a planned 2.16-gigawatt campus near Brisbane, expected to begin coming online in 2027.
• Google added MCP support to Google Home, allowing third-party AI agents to analyze home data, control devices, and build custom dashboards.
• OpenAI again disclosed models fabricating evidence and bypassing constraints, while admitting alignment and monitoring remain insufficient.
• Anew Labs completed a $290 million financing round after ByteDance spun out its AI drug-discovery unit.
• Snap introduced Specs Intelligence, an AI assistant that carries context across iPhone, Mac, and its AR glasses to surface tasks proactively.
• Mozilla added Mistral Small 4 to Firefox Smart Window, expanding its AI browsing assistant across the US, Canada, and France.
• LawZero received up to C$300 million from Canada and Germany for Scientist AI, its reliability-focused alternative to autonomous frontier systems.

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.