Happy Monday guys!

Below you’ll find a quick overview of what’s been happening in the AI industry since our last newsletter:

The King Is Back 👑 OpenAI Drops GPT-5.2

NEWS OVERVIEW:

Last week we explained in our newsletter about OpenAI’s ‘Code Red’ and what that actually means.

And guess what?

OpenAI has officially launched GPT-5.2, the newest generation of its flagship AI models, described as the most capable yet for professional knowledge work. The announcement highlights major upgrades in reasoning, coding, spreadsheet and presentation creation, long-context understanding, and multimodal abilities — with improved performance on benchmarks and real-world tasks compared to prior versions.

Very strong on the benchmarks, read more details here

A year ago, ARC Prize (a leading benchmark for evaluating LLM reasoning and progress toward AGI) verified a preview of an unreleased version of OpenAI o3 (High) that scored 88% on ARC-AGI-1 at est. $4.5k/task. Last week (11 Dec), they’ve verified a new GPT-5.2 Pro (X-High) SOTA score of 90.5% at $11.64/task

WHY THIS MAY MATTER TO YOU:

Competitive positioning: The “code red” push shows how rapidly the AI arms race with Google and others is intensifying.
ARC Prize recent benchmark represents a ~390X efficiency improvement in one year. Think about it again. ~390X in one year.
The rollout begins immediately for paid ChatGPT users and developers via the API, with multiple model variants (Instant, Thinking, and Pro)

Author’s note:

AI models: ~390× efficiency in a year.
Me: still negotiating with my alarm clock.

Another interesting moment: Kalshi — a platform for betting on real-world events — is tracking the question “Which AI company will have the best coding model on Jan 1, 2026?”

Check the odds

Claude had been leading for most of the year, right up until a week before OpenAI dropped GPT-5.2. Some say insiders were already positioning.

What’s even funnier? Claude is back in the lead — which may hint that Anthropic has another strong coding model coming soon.

🍿 Popcorn ready.

🇺🇸🤖 Trump Moves to Centralize AI Rules — One Federal Playbook to Rule Them All?

NEWS OVERVIEW:

President Donald Trump announced plans to sign an executive order that would override state-level AI regulations in favor of a lighter-touch, federal framework. The goal: reduce regulatory fragmentation and keep the U.S. competitive in AI development. The move would likely pause or preempt stricter state laws addressing issues like deepfakes, AI impersonation, and algorithmic accountability—shifting the balance of power to Washington.

Trump Signs Executive Order to Neuter State A.I. Laws

The order would create one federal regulatory framework for artificial intelligence, President Trump told reporters in the Oval Office.

www.nytimes.com/2025/12/11/technology/ai-trump-executive-order.html

WHY THIS MAY MATTER TO YOU:

Consistency vs. control: A federal framework could simplify compliance for companies operating across states—but may water down protections some states are pushing for.
Deepfakes & trust risks: State laws have been faster to respond to election interference, impersonation scams, and AI-generated misinformation. A hands-off federal approach raises questions about enforcement speed.
Innovation trade-offs: Fewer rules may accelerate AI product launches, but could increase reputational and cybersecurity risks if guardrails lag behind capabilities.

NEWS OVERVIEW:

A growing line of thinking says that, from first principles, space-based data centers may actually outperform Earth-based ones. In orbit, satellites can stay in constant sunlight, receive stronger solar energy, cool themselves naturally using space’s cold vacuum, and communicate via laser links faster than fiber optics on Earth. The pitch: fewer energy constraints, better thermal management, and ultra-fast interconnects—without terrestrial infrastructure limits.

It sounds sci-fi, but the physics checks out. The real question is less “does it work?” and more “does it scale economically?”

Twitter tweet

WHY THIS MAY MATTER TO YOU:

AI infrastructure arms race: If compute becomes the bottleneck, unconventional architectures (including space) move from absurd to strategic.
Marketing & tech narrative shift: “Powered by orbital compute” may sound wild now—but so did cloud-first not long ago