The web says no to agents, Nvidia dictates its suppliers' factory, Bill Gross gives away 1.2 million ideas

Three signals this Monday. The internet is starting to say no to the AI agents that were supposed to work while you sleep. The most powerful company in AI now tells its own suppliers how to manufacture. And the man who built a career on ideas just gave away 1.2 million of them, for free.

Signal one: the web closes its doors to agents

The always-on agent was the industry's big bet. OpenAI launched Dots at its DevDay on September 29. Meta's Muse was downloaded millions of times. xAI's Grok Bot answers mentions. The promise: agents that work while you sleep, that book, buy and handle chores.

The Verge reported on Friday that the rest of the internet is not cooperating. Users of Muse, Grok, Dots and Instinct say their agents are getting blocked on major sites:

• Yelp told TechCrunch it only allows agents that pay for its data licensing program.
• Delta Air Lines prevents agents from booking tickets for users.
• JPMorgan Chase and Charles Schwab confirmed they block agents on customer accounts, for security reasons.
• Some users say Cars.com and Amtrak no longer work with their agents. Neither company confirmed it.

The commercial logic

There is nothing subtle about it. Last year Amazon banned Perplexity's agentic shopping feature and sued the start-up. A judge ordered Perplexity to stop. An appeals court overturned that decision in August. Amazon has since filed an amended complaint, and told Meta that Muse's continued access violates its terms of use. The first piece in this series already described Amazon blocking Muse.

Perplexity's reading of Amazon's motives is blunt: agents do not look at ads. Meta makes the opposite case: refusing a personal agent means refusing the customer behind it. Meta has shopping partnerships with Walmart, Gap, Sephora and Best Buy. Shopify has opened its network of stores to agents.

So two internets are forming: one that negotiates with agents, and one that shuts them out.

Why it matters

The AI giants' entire agent strategy assumes the web will cooperate. It will cooperate on its own terms: for payment, after negotiation, and once someone settles who is liable when a credentialed agent gets it wrong.

Schwab chief executive Rick Wurster said on October 16 that his firm blocks dozens of large AI agents every week, naming Muse, and raised exactly that liability question. JPMorgan's Jamie Dimon warned about rogue agents on the October 13 earnings call.

The agent is your stand-in. It logs in, fills the cart, pays. Sites now have to decide whether that stand-in is a customer or an intruder. Most answer with a price list or a security policy.

What changes

For brands, the distribution math is being rewritten. If agents become the interface between your customer and your storefront, your relationship with agent platforms becomes a commercial negotiation: licensing fees, partnerships, preferred placement. Yelp just provided the template: charge the agent for the data.

I see three possible postures for a brand, and each is a strategic choice:

1. Open, like Shopify, to be where agents buy.
2. Monetize, like Yelp, when the data is the asset.
3. Close, like the banks, when risk outweighs the rest.

The only wrong choice is not choosing, and discovering in your traffic logs that the decision was made by default.

For those building on agents, the lesson is harder. A capable assistant still needs somewhere to work, and the map of those places is being drawn by lawyers as much as engineers. The agentic web will be filtered, licensed and commercially negotiated. Plan accordingly.

Signal two: Nvidia dictates its suppliers' factory

Samsung Electronics will use TSMC's 3-nanometer process to make the base dies for Nvidia's next generation of custom memory. The Seoul Economic Daily reported it, and Reuters picked it up.

Why it is unusual: Samsung and TSMC are rivals in chipmaking. It is the first time Samsung has worked with the Taiwanese foundry at this stage of memory manufacturing. And it was not Samsung's idea. Nvidia asked its three memory suppliers, Samsung, SK Hynix and Micron, to use TSMC's process for the base dies and sixth-generation DRAM for the core dies. Samsung agreed after an internal review.

The memory in question

It is Nvidia's NVHBM, unveiled in August. HBM, high-bandwidth memory, is a stack of memory chips placed as close as possible to the processor, essential to feed AI computation. NVHBM builds Nvidia's custom memory controller into the base die, with the stated goal of standardizing implementation across suppliers.

Standardize is the polite word. The accurate word is dictate. The customer now specifies its suppliers' manufacturing process. Samsung keeps in-house production for standard memory and other customers, and is reportedly assessing whether its own foundry could step in if TSMC runs short of capacity. That last detail says everything about the balance of power: plan B is asking permission later.

Why it matters

AI runs on memory, and memory is scarce. Analysts expect a high-bandwidth memory shortage of about 20% this year, lasting until 2028. Every product roadmap, every data center build, every model training run depends on this chain. When the dominant buyer starts dictating its suppliers' manufacturing processes, you see what market power looks like when it stops pretending.

What changes

TSMC wins whichever memory maker delivers. That is the position every strategist should study. In a gold rush, the sure winner rents the shovels to everyone, while the loudest miners make the headlines.

For companies preparing AI products, the practical lesson is cost and timing. Your AI margins are physical: foundry capacity, memory supply, power contracts. Price them in now, because the company at the top of the stack is already doing it for you.

Signal three: the idea man gives his ideas away

Bill Gross launched Idealab.ai. He founded the Idealab incubator in 1996 and backed more than 175 companies. The new platform is a free catalog of more than 1.2 million AI-generated business ideas.

He ran about 17 million sources (news, expired patents, SEC filings, complaints posted on Reddit) through a system built around his 31 methods for spotting opportunities. According to FounderBrew, 36 million concepts were winnowed down to 1.2 million. Cloudflare chief executive Matthew Prince pledged 100,000 dollars in credits to each start-up built on the platform. The stated ambition: a million new businesses and a hundred million jobs.

The quote

It is the story. "My whole life has been coming up with ideas myself, and now AI can do that better than me." The man whose entire career was the idea looks at the machine and concedes the one thing that made him famous. He adds that every business decision is still made by the human. The human remains the conductor. The machine handles the mechanics.

Why it matters

For thirty years, the scarcest ingredient in entrepreneurship was supposed to be the idea. That was the mythology: a big idea, a garage, a unicorn. Gross just priced the idea at zero. One point two million ideas, free, searchable by your skills and your capital.

If he is right that the machine outperforms the idea man, then the start-up scarcity story was wrong from the start. The real scarcities were always elsewhere: judgment, execution, distribution, and the unglamorous ability to keep going when nothing works.

What changes for brands

The question deserves reflection. If idea generation becomes a commodity, what is your real edge?

Brainstorming has become a commodity. What remains: taste, trust, the distribution you already own, and execution nobody can copy from a prompt.

Gross removed about 100,000 potentially harmful ideas before launch. That detail tells the other half of the story. Judgment remains the craft. The craft has changed shape: it moves from having the idea to knowing which one deserves to exist.

It is exactly what I argue in my own work: diagnosis before prescription. An idea without an accurate reading of the situation remains a hypothesis. Gross's million ideas make that reading more valuable, because choosing now happens amid abundance.

The thread of the day

The web closes its doors to your agent, and the entry price is being negotiated. AI's most powerful buyer tells its suppliers which factory to use. And the idea man gives away 1.2 million ideas, because the idea was never the scarce part.

Power is shifting toward those who control access, factories and judgment. A brand needs to know which of the three it owns.

The noise filtered. The signal kept.