🛎️ The Free Internet Dead

Good Morning, AI Enthusiasts!

The bill for the AI era is being passed around, and almost everyone in the room is paying it except the model.

TOKEN

The Free Internet Dead

👀 What’s happening: Meta launched global paid tiers Wednesday under one Meta One brand for Instagram and Facebook with two Meta AI tiers at $7.99 and $19.99, a creator pro tier at $49.99. The consumer tiers unlock anonymous Story viewing, deeper analytics, custom themes and privacy controls; the AI tiers buy faster responses, higher limits, and advanced reasoning. The rollout lands a week after Meta cut 8,000 staff and committed $125 to $145 billion to AI infrastructure.

🌍 How this hits reality: X and Snap went subscription years ago because their ad businesses were never strong enough alone. Meta is the opposite case, its ad machine still works, three billion users monetized at industry-best margins. Subscriptions arrive anyway, because ads built for a CPM world cannot underwrite a $145 billion AI capex bill. So the strongest ad business on Earth concedes that ads alone no longer cover the next decade.

🛎️ Key takeaway: Snap and X went paid because ads were never enough, Meta is going paid because ads at full strength still cannot pay AI's compute bill, and that gap is the real price tag of the AI era.

TOGETHER WITH BOLT

You know the tool your team needs but no one has time to build? Open Bolt.new, describe it in plain English, and ship the first version this weekend. Real code. Real deployment. No engineer required.

Founders are spinning up CRMs, dashboards, and client portals between coffee and dinner.

RESEARCH

Erdős Now Look Like an Intern Task

👀 What’s happening: Anthropic researcher Levent Alpoge spent a weekend running Claude Mythos against the Erdős unit-distance conjecture, air-gapped, multiple instances, no access to OpenAI's published proof. The model produced an independent disproof shorter and cleaner than OpenAI's 125-page chain of thought, reaching for a class field tower from algebraic number theory the human tradition never touched. The same week, OpenAI cracked the 80-year wall first and DeepMind's AlphaProof Nexus cleared nine more Erdős problems at a few hundred dollars each.

🌍 How this hits reality: Erdős left over a thousand open problems, each one a small monument to how hard human mathematicians had to think. Anthropic just produced a cleaner proof than OpenAI's over a weekend, and in the same seven days the other two frontier labs walked through the same wall by their own routes. None of the labs coordinated. The convergence is the signal. When Claude, GPT and Gemini independently treat an 80-year open problem as a weekend exercise, the bottleneck stops being the problem and starts being the mathematician standing in front of it.

🛎️ Key takeaway: Erdős spent a lifetime building a thousand walls meant to outlast generations, and the only open question left is what a research mathematician is supposed to do on Monday.

VIDEO

The AI Video Got Labeled

👀 What’s happening: YouTube will auto-detect and label "significant photorealistic AI" video this month, ending two years of creator self-disclosure. Labels move from the buried description to directly below the player and overlaid on Shorts. Anything made with YouTube's own Veo or Dream Screen, or tagged with C2PA, carries a permanent label creators cannot remove.

🌍 How this hits reality: Audiences flinch at the "AI" stamp, so don't volunteer it. Self-disclosure was a polite fiction nobody enforced, and the smart move was a vague description and no checkbox. That dodge just got walled off by Google. Channels built on calling the synthetic parts "edited" now wear a label they cannot peel, on a surface viewers actually see. Stock farms, faceless explainers, AI music grinders all get retagged in public.

🛎️ Key takeaway: Audience disgust was the last unpaid bill on synthetic video, creators were ducking it by staying quiet, and the platform just stapled the receipt to their forehead.

TOKEN

OpenAI Cheated Customers

👀 What’s happening: GPT-5.5 users caught the swap this week. A user reported his ChatGPT Extended Thinking session silently dropped to Instant after two hours, while the UI label never moved. OpenAI's help doc admits Plus accounts get rerouted to mini after 160 messages per three hours, no popup, no badge. Pro at $200 a month faces the same throttling on Heavy thinking. A February Codex trace already caught a Pro 5.3 request returning 5.2 base.

🌍 How this hits reality: This is the lab playbook, not an OpenAI accident. Anthropic got caught pulling the same trick on Opus 4.6, quietly routing premium calls to a smaller sibling under load. Frontier compute is the most expensive line on the balance sheet, subscriptions are the cheapest revenue these companies sell, and the math has exactly one answer. Paying customers fund the GPU bill, then get served whatever the load balancer can spare, with the badge left on for decoration.

🛎️ Key takeaway: Every frontier lab treats its paying subscribers as the shock absorber for its compute bill, Anthropic did it with Opus, OpenAI is doing it with 5.5, and the only thing premium about the tier is the invoice.

DAILY TL;DR

• Robinhood opened its platform to AI agents, letting users authorize automated trading and money actions.
• Cognition raised a new Series D to push Devin into more enterprise coding workflows.
• Amazon introduced agentic shopping tools for retailers, pushing autonomous commerce deeper into online stores.
• Chan Zuckerberg Initiative unveiled an AI world model for drug discovery, targeting faster biological simulation.
• Valve raised Steam Deck OLED prices as AI-driven memory demand pushes hardware costs higher.
• Salesforce reported stronger guidance as Agentforce demand becomes a bigger part of its AI growth story.
• Wix plans to cut about 1,000 jobs as losses, costs, and AI pressure rise, while continuing to bet on Base44 and its own AI model.
• SK Hynix topped $1 trillion for the first time as AI memory demand and price hikes lifted Korean chip stocks.

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.