The AI Inflection Point: Anthropic's Near-Trillion IPO, Bots Outnumber Us Online, and a Vaccine Designed by AI
Five stories that made us stop and say — wait, that already happened?
We say this a lot, but this week genuinely felt different.
In the span of seven days, the company behind Claude filed for one of the largest IPOs in history, bots officially outnumbered humans on the internet for the first time ever, AI designed a vaccine that went into human trials, and one of the most celebrated filmmakers alive quietly handed his storyboards to a machine. Oh — and Nvidia decided it was time to come back to your laptop.
There's a lot to unpack. Let's get into it.
💰 The Money Moves That Are Reshaping AI
Anthropic files for a $965 billion IPO
If you've been following the AI race from the outside, this week gave you a door in. Anthropic — the lab behind Claude — filed a confidential draft S-1 with the SEC, setting up what could be a public listing as early as this fall. The valuation has rocketed from $380 billion to $965 billion in a matter of months, supported by a $65 billion Series H raise and a revenue run rate that grew from $10 billion to roughly $47 billion in a single year.
What makes this remarkable isn't just the number. It's that frontier AI has been an insiders-only club — venture funds and strategic investors, not ordinary portfolios. An Anthropic IPO would change that. For the first time, the people using these tools every day could own a piece of them.
Warren Buffett bets big on AI infrastructure
At the same time, Alphabet announced an $80 billion capital raise — the largest in US corporate history — to fund AI compute. Berkshire Hathaway anchored it with $10 billion. When the most famously patient investor in the world makes a move like this, it's worth paying attention. The infrastructure to run AI at scale isn't cheap, and the biggest players are clearly in a race to build it first.
🌐 The Internet Quietly Crossed a Threshold
Bots now outnumber humans online — and it happened ahead of schedule
Cloudflare's CEO expected this moment in 2027. It happened this week.
For the first time in the web's history, bot traffic has overtaken human traffic. Cloudflare monitors roughly one in five websites globally, which makes its data a fair proxy for the whole internet. Bots sat at around 20% of requests before the AI era. They've since grown eight times faster than human activity over the past year — driven not by old-school scrapers, but by AI agents taking over tasks people used to handle themselves.
Here's the part worth sitting with: a single AI agent can visit 5,000 websites in the time a human visits five. It compares prices, researches topics, and books trips on your behalf. That's genuinely useful. But it also means the sites that get seen, the reviews that get surfaced, and the options that get ranked are increasingly filtered through a machine's judgment rather than your own.
What this means for you: If you're a founder, marketer, or legal professional building an online presence, you're not just optimizing for human readers anymore. The audience is increasingly AI agents deciding what's worth surfacing to humans. That changes the game for content, SEO, and discoverability in ways we're only beginning to understand.
🚀 AI as a Creative and Building Tool
OpenAI ships prompt-to-app
OpenAI quietly rolled out a feature called Sites inside Codex this week. The pitch is simple: describe an app in one sentence, and it builds it, hosts it, and hands you a live URL — complete with databases, file uploads, and sign-in. No developer, no setup.
This idea isn't new. Lovable crossed $400 million in revenue and Replit reached a $9 billion valuation selling it. What's different now is the distribution. Codex already lives inside ChatGPT, which millions of teams pay for every day. OpenAI isn't inventing this market — it's walking into one with a very large existing audience. Currently available on Business and Enterprise plans, with broader access on the way.
Martin Scorsese's storyboards are now AI-assisted
This one landed differently than a lot of the "AI in Hollywood" stories we usually see. Martin Scorsese — 50+ years of filmmaking, hands on every frame — has joined Black Forest Labs as a partner and advisor, and is using their FLUX model to plan shots for his current production. He's not replacing actors with AI or generating anything audiences will see. He's sketching scenes for his cast and crew, a task he's done by hand for most of his career. He described it as cutting setup time while keeping his visual judgment intact.
It matters because of who it is. If someone with Scorsese's relationship to craft is finding a genuine use for this at the planning stage, the conversation about AI and creativity just got a lot more nuanced.
Google Flow now plans entire films
Alongside the Scorsese news, Google's Flow tool has expanded to support full production planning — not just individual clips, but coherent story structure across a whole film. For context, Flow was already generating impressive short-form video. The move into long-form production planning signals that the tools for AI-assisted filmmaking are maturing quickly.
🔬 The Story That Quietly Changed Everything
An AI-designed vaccine is now in human trials
We almost buried this one, which would have been a mistake.
A team at the University of Cambridge fed an AI genetic data from dozens of coronavirus variants — including strains still circulating in animals. The model identified patterns that human researchers hadn't prioritized and designed a vaccine targeting not just current strains, but the broader coronavirus family. That vaccine has now entered human trials. Safety testing covered 39 people initially; a study of around 200 is underway.
The core of the vaccine was designed entirely by AI — a first for the technology in human clinical trials. We're still at an early stage, and efficacy data will take time. But if it works, this is the proof-of-concept that AI can do something in medicine that humans couldn't do fast enough on their own.
⚡ One More: Nvidia Returns to Your Laptop
RTX Spark brings data-center AI power to Windows PCs
Nvidia has been the engine behind almost all large-scale AI training for years. Now it wants to be inside your laptop. The RTX Spark chip combines a 20-core Arm CPU with a Blackwell GPU on a single piece of silicon, sharing up to 128GB of memory and delivering what Nvidia claims is a full petaflop of AI compute. The first laptops arrive this fall from Dell, Lenovo, and Microsoft's Surface line.
The practical implication: local AI that doesn't need a server. An agent that can run in the background, sort your email, or process documents without sending your data to the cloud. That matters for privacy, for speed, and for the growing number of people who want AI capabilities without the dependency on external infrastructure.
A Thought to Leave You With
Anthropic — the same lab warning this week that recursive self-improvement could be two years away — is simultaneously building tools to detect and potentially pause that process. Claude now writes over 80% of Anthropic's own merged code. One of its engineers reportedly hasn't typed a single line of code himself in five months.
The people closest to this technology are the ones raising the alarm most clearly. That's not a contradiction. It's a sign that we're at a moment where the decisions made in the next few years will matter enormously — and where transparency, governance, and human oversight aren't abstract ideals. They're urgent engineering and policy problems.
That's exactly what we talk about on AI Insights every week.
Enjoyed this edition? Come find us at aiinsightsnetwork.com — and if a story here sparked a question or a perspective you want to dig into further, that's what the podcast is for. Fleur and Grainne break this down every week, with guests who are actually in the rooms where these decisions are being made.
Until next week — stay curious, stay skeptical, and stay human.
Fleur & GrainneAI Insights
You're receiving this because you subscribed to the AI Insights newsletter. To unsubscribe, click here.
WORD COUNT: ~900 words of body copy ESTIMATED READ TIME: 4–5 minutes