AI Arms Race Volume: 2

Weekly AI Signals — Curated by Builders, for Builders

Every week, we round up the most relevant developments in AI, tech, and the business strategies shaping them. Born from internal conversations among senior engineers, product leads, and AI researchers, this digest simplifies the noise into clear signals. Authored by Brod Justice and Ryan McClure, these summaries are built for real practitioners—people shipping code, scaling platforms, and thinking past the hype.

Volume 2

The Big Gets the Buzz. The Small Gets the Win.

1. The Size Trap

The Register put it bluntly: the biggest AI models might be the dumbest investment in the space. While OpenAI basks in the hype, it’s the small and mid-sized models that are transforming workflows with speed and affordability. (Even GPT-4o, which we covered last week, is now facing scrutiny for its ballooning inference costs.)

2. China’s GLM-4.5: Open, Powerful, Ignored

Zhipu AI quietly released the GLM-4.5 Series, an open-source LLM now ranking among the top 5 in the world. Why no buzz? It’s Chinese, free, and not backed by Silicon Valley PR.

3. $500B Down a Black Hole

SoftBank’s Stargate project with OpenAI was supposed to be AI’s moonshot. According to the WSJ, it’s already floundering. No data center. No launch. No agreement between Altman and Son. No surprise.

4. 15% Vanished Traffic: AI’s Zero-Click Apocalypse

A growing number of news publishers report sharp traffic drops. The BBC asks the existential question: can the web survive AI summaries? AI is feeding on the web — and not giving back.

5. Copyright Collapse: From Germany with Alarm

In Süddeutsche Zeitung and Tagesschau, Andrian Kreye frames it plainly: copyright is the first casualty in the USA–China AI arms race. The race now is to defend copyright before it vanishes.

6. Pick a Number — AI Always Chooses 27

Want to break your trust in LLMs? Ask it: Pick a number between 1 and 50. Most say 27. It’s not just funny — it’s revealing. Biases aren’t bugs. They’re baked in. More here on the Business Landing site.

7. Haircut AGI: The Zuckerberg Snark

One user nailed it: “Zuckerberg says Meta AI is about to take all our jobs. I’ll believe in superintelligent Meta AI when it cuts his hair.” This sarcastic zinger went viral — and it captured the current AI hype fatigue perfectly. Humor is the last defense against tech delusion.

8. AI Summaries Are Gutting News Traffic

A Guardian study finds that AI summaries are causing a dramatic collapse in audience reach for online publishers. Google and OpenAI are summarizing away the media business model.

9. From Napster to Spotify — For Search

We Are Social says AI is doing to search what Napster did to music. Search is dead. Curation is coming. SEO isn’t declining. It’s decomposing.

10. Trump “Abolishes” Copyright

In the same SZ article mentioned above, Trump’s new AI policy essentially kills copyright enforcement in the U.S. Misinformation just became protected speech.

11. Psychologists Warn: Stop Confiding in ChatGPT

Experts are sounding the alarm: people are emotionally bonding with chatbots. The bots aren’t sentient — but users are treating them like therapists. We don’t need AGI to break down. We’ll do it ourselves.

12. Stanford: Small Models Take the Lead

The 2025 AI Index confirms: model size is shrinking fast — yet performance keeps rising. 3B parameters is the new 300B. The real disruption isn’t intelligence. It’s efficiency.

13. Mistral Medium 3: The Unsung Hero

Mistral’s Medium model is outperforming bigger rivals at a fraction of the cost. Think Claude Sonnet 3.5 quality — but far cheaper and more deployable. It’s not how big your model is. It’s how smart it runs.

14. Meta Reverses on Open Source

Meta is walking back its open-source commitments. Despite years of “open” AI evangelism, Meta is walking it back. No more open weights for top-tier models. Open is for PR. Closed is for competition.

15. China’s Global AI Governance Pitch

At WAIC in Shanghai, China introduced a UN-aligned AI policy focused on real-world deployments and sovereign infrastructure. China’s play isn’t just tech. It’s global influence.

16. Meta’s Secret Superintelligence Lab

Zuckerberg’s mystery AI lab is staffing up with elite researchers and billions in backing. AGI? Meta wants in — and they’re not telling us how.

17. Cuban: Ban Ads in AI Agents

Mark Cuban wants the U.S. to crack down on ads inside AI. Referral bias, hidden sponsorships — it’s the new frontier of misinformation. We fixed Google search, just in time to break it again.

18. Google Faces EU Complaint Over AI Summaries

European publishers have filed an antitrust complaint, alleging Google’s AI Overviews are devastating their web traffic. The case for fair use just got murkier.

19. Elon’s xAI Snubs EU AI Pact

Musk’s xAI is refusing most of the EU AI code, calling it “detrimental to innovation.” He’s only signing the safety portion. Safety, yes. Regulation, no thanks.

20. U.S. Gov Divided on China Chip Sales

Washington remains split over AI chip sales to China, with defense hawks clashing with economic pragmatists. Even in politics, AI is reshaping trade logic.

21. Sustainable AI: Growth Has Limits

A recent arXiv paper argues AI scaling is hitting limits — economically, environmentally, and socially. The cost of big models may soon outweigh their value. We’ve built faster. Now we need to build wiser.