In partnership with
🌟 Editor's Note: Recapping the AI landscape from 11/11/25 - 11/17/25.
🎇✅ Welcoming Thoughts
Welcome to the 19th edition of NoahonAI.
What’s included: company moves, a weekly winner, AI industry impacts, practical use cases, and more.
Michael Burry bet against AI then decided to close his firm 3 days later.
Jeff Bezos is in on the AI race! More in Impact Industry.
Fun interview today with an old baseball teammate of mine who just started a cool company out in Austin.
Perplexity was voted most likely to fail at a major AI conference.
Cursor just raised $2.3B at a $29B valuation. AI coding tools continue to be one of the most dominant AI use cases.
Interesting article here speculating on OpenAI’s cash burn.
An AI startup that sets up post-death conversational avatars got absolutely crushed on socials this week.
Em-dashes may finally be out of GPT if you add the requirement in GPT customization settings (see Issue 17 PUC).
NVIDIA earnings this week will be very interesting.
I need to set up an AI agent so I can “talk to” all of my newsletters at once and come up with more ideas.
Seeing more humanoids and applied robotics come out every day. It’s awesome.
Let’s get started—plenty to cover this week.
👑 This Week’s Winner: Anthropic // Claude
Anthropic is this week’s winner → They started it off with a “minor” partnership in today’s AI economy, with Fluidstack, it centered around a $50B investment towards building data centers across the US, primarily in NY and Texas.
More from Anthropic:
Disrupted AI-Driven Cyber Espionage: Anthropic says it detected and stopped a state-linked cyber-espionage campaign that misused Claude Code against ~30 global targets.
Showcased Claudius Agent: Anthropic revealed “Claudius,” its autonomous vending-machine CEO experiment, giving the public a look at real-world autonomous-agent behavior, failure modes and guardrail testing.
Expanded Government Footprint: Maryland announced a statewide partnership to deploy Claude across public-service agencies (benefits, housing, child-poverty work), marking one of the strongest government endorsements of an LLM to date.
Anthropic also released its own benchmark to test political bias in LLM’s. Sonnet 4.5 scored a 94% in this area and sits just behind Grok and Gemini. Overall, a strong week for Anthropic headlined by the new data center deal.
From Top to Bottom: Open AI, Google Gemini, xAI, Meta AI, Anthropic, NVIDIA.
⬇️ The Rest of the Field
Who’s moving, who’s stalling, and who’s climbing: Ordered by production this week.
⚪️ NVIDIA
Strong MLPerf Training: NVIDIA GPUs powered many of the top MLPerf Training results, reinforcing their lead in AI data-center hardware.
$1.4B in Recent Acquisitions: NVIDIA has spent about $1.4B buying AI-infrastructure companies like Run:ai, OctoAI, Shoreline and Deci to strengthen its software and tooling stack. Stealth moves. Infrastructure companies.
Arm–NVIDIA Link Deal: Arm’s Neoverse CPUs will now connect directly to NVIDIA GPUs through NVLink Fusion, making it easier for cloud providers to pair custom Arm chips with NVIDIA accelerators. Making the right moves.
🟢 OpenAI // ChatGPT
GPT-5.1 Released: OpenAI launched GPT-5.1 with major speed gains, also includes better tool-calling performance, extended prompt-cache, faster code edits, and improved reasoning. Solid update, did not address context window (memory), which is my only big issue with ChatGPT.
Group Chat Feature: OpenAI rolled out multi-user group chats (1–20 participants) in select countries, enabling multiple people to use the same ChatGPT conversation. This is cool and a smart move. Not in US yet.
Copyright + Data Fight: OpenAI is challenging a court order to hand over 20 million chat logs in a NYT copyright case, while also facing a Munich ruling that GPT unlawfully reproduced copyrighted song lyrics. Essentially all of these LLM’s trained on copyright data in some capacity so this isn’t going away. Although privacy concerns may halt any conversation sharing.
🔴 xAI // Grok
Grok 5 Coming Q1 2026: Elon confirmed Grok 5 will launch in early 2026 as a 6-trillion-parameter multimodal model handling text, images, video, and audio with a new design for higher efficiency. Intrigued by the video reading/intake potential here. Something most models can’t do.
Grok 4.1: xAI rolled out Grok 4.1 with major improvements to personality, creativity, and emotional awareness. 65% of people preferred 4.1 over the prior model in blank tests.
Grok Code in Development: xAI is building Grok Code Remote, a cloud-based coding environment with repo and PR integration, positioning it to compete with OpenAI’s Codex and Claude Code. There is a huge market for these tools right now and if done well it could be a major boost for xAI.
🟣 Google // Gemini
Gemini 3 + Nano Banana 2: Reports suggest Google will drop Gemini 3.0 Pro and Nano Banana 2 this week, with upgrades to image generation, faster rendering and better prompt accuracy. Regarded as what should be an impressive launch. Very excited for Nano Banana 2. Nano Banana 1 is the best AI image model out right now.
WeatherNext 2 Released: DeepMind launched WeatherNext 2, an 8x faster forecasting model using a new Generative Network. Cool to see Google’s advancements outside of general LLM / Media areas.
New Gemini Certifications: Google rolled out three free Gemini AI certifications for educators and students, offering multi-language exams and digital credentials. Nice. Learn AI!
🔵 Meta // Meta AI
AI Godfather Leaves Meta: Yann LeCun is departing Meta to build a new company focused on world-model AI, reflecting his belief that LLMs alone can’t reach AGI and signaling a philosophical shift away from Meta’s research roots. Yann started AI at Meta in 2013. Interesting philosophy that goes against today’s leading voices.
Required AI Usage: Starting in 2026, Meta will bake “AI-driven impact” directly into performance reviews, making AI adoption and productivity gains a required part of every role. Good move.
Marketplace AI: Meta is upgrading Marketplace with AI-powered suggested questions and detailed vehicle insights, pushing Marketplace deeper into AI-driven shopping experiences. Cool.
🏢 Impact Industries 🚑
Enterprise // Project Prometheus
Jeff Bezos is creating Project Prometheus, where he will act as Co-CEO. It is an effort to build a next-generation AI supercomputer designed to learn and reason more like a human brain. Instead of scaling today’s models, the project explores new computing architectures that could make AI more adaptable, efficient, and capable of long-term reasoning. Insiders describe Prometheus as Bezos’ boldest bet yet, a move to push beyond brute-force compute toward a more intelligent kind of AI.
Medical // Skin Cancer Detection
Researchers at Incheon National University developed an AI system that detects melanoma with 94.5% accuracy, nearly doubling precision over existing methods. The multimodal model analyzes both skin images and patient metadata, allowing for earlier and more reliable diagnosis. By combining visual and contextual cues, the system could significantly improve cancer detection and treatment planning, a major step toward AI-powered dermatology that improves outcomes through faster intervention.
🎙 Weekly Interview: 10 Minutes With James Hamilton
James Hamilton
🏠 Background: James Hamilton is a software engineer and founder based in Austin. He’s an Ohio State grad who moved from engineering into the creator-growth world, becoming known as “The Clipping Guy.”
💼 Work: James is the founder of Dubble, a platform that helps creators grow through large-scale clipping campaigns. Before this, he built Local Brew and worked as a software engineer at eFuse and Outern.
🚀 Quote: “AI just lets you move faster. What used to take me a month to build, I can now do in two days.”
10-Minute Interview — James Hamilton (Founder, Dubble)
1. Tell us about yourself and what you’re building.
James: I’m James Hamilton, originally from Ohio and now in Austin. I studied Computer Science at Ohio State, worked as a software engineer, and now I’m building Dubble — a platform that helps creators grow through large-scale short-form clipping campaigns. We turn long-form videos into human-edited clips for TikTok, Instagram, and YouTube.
2. How does Dubble use AI today, and where is AI clipping going?
James: AI clipping tools are improving, but they still miss real creative moments—hooks, emotional beats, payoffs. You still need humans in the loop. Our clippers use AI to assist, but the creativity is human-led. On the engineering side, AI helps us ship features way faster than before.
3. What’s your take on humanoid robotics and automation?
James: Robotics is super exciting—especially in manufacturing. Figure is doing impressive stuff, and Tesla’s Optimus is wild. Controlled environments make early adoption easier. I don’t follow robotics deeply, but it’s obvious it’ll be a massive part of the future.
4. What excites you most about the next 3–5 years of AI?
James: Content creation will change the fastest. AI lets creators iterate quickly—almost “speedrun” the creative process. You’ll find content-market fit faster. There’s a world where AI avatars take over feeds, and a world where AI just boosts human creativity. We’ll see which path wins.
5. What’s your view on the surge of AI-generated content?
James: Some of it is hilarious, and if people enjoy it, great—the market decides. The only real issue is when someone’s art is copied or stolen. At Dubble, we stay focused on human-edited content.
👨💻 Practical Use Case: Claude Artifacts
Difficulty: Basic
Claude Artifacts are one of the most interesting product ideas launched by Anthropic this year. Think of them like Claude’s version of Custom GPTs — mini-apps that live directly inside a chat. Instead of just returning text, Claude can generate fully interactive tools: dashboards, study helpers, utilities, and even simple games.
What makes Artifacts unique is the built-in transparency.
Claude shows you the exact prompt it used to create the artifact inside the chat window. This means you can see:
How the artifact was constructed
How Anthropic writes its own internal prompts
How other users are structuring advanced instructions
How to tweak or rebuild the artifact yourself
You can also create your own artifacts. This can be helpful for repetitive tasks and even act as a simple exercise to get you more comfortable working with AI.
Examples of Artifacts people are already building:
QR Code Generator
Language Tutor
Minigames (puzzles, blackjack, Snake, etc.)
If you want to try it for yourself:
Open Claude
Click “Artifacts” in the top left & choose one
Inspect the prompt it exposes
Modify the instructions
Regenerate until it works exactly how you want
I’d say artifacts, similar to custom GPT’s (in app), are nice-to-have’s and useful. In most cases they aren’t game-changing technologies, those are often reserved for apps built separate from LLM’s, but I do like the window they give into the prompt writing of the creators, and the mockup speed for testing ideas.
Learn more on how to create your own artifacts here ⬇️
📲 Startup Spotlight
Dubble
The Problem: Brands and creators spend huge budgets on ads and influencer deals that often miss their mark. Getting consistent organic reach, especially from short‑form video, is tough, but valuable.
The Solution: Dubble connects brands with an “army” of verified micro‑creators who create and share clip‑based content across hundreds of accounts. Brands upload their long‑form content, set guidelines, and Dubble handles everything: clip creation, posting, tracking views, and iterating based on performance. The result: organic mass distribution for a fraction of traditional ad cost.
The Backstory: Founded by Ohio native and Ohio State Graduate James Hamilton, Dubble is built to proactively align marketing with current consumer trends. Short-form content is dominating audience screen time, and Dubble allows for brands to take full advantage of the current landscape.
My Thoughts: Short-form is key. Tiktok, YT Shorts, IG / FB Reels, it’s everywhere. From my experience it’s also very time-consuming to take long-form video and turn it into the right medium that will achieve some form of virality. I like what James and Co. are doing here, check it out!
“It’s not likely you’ll lose a job to AI. You’re going to lose the job to somebody who uses AI”
- Jensen Huang | NVIDIA CEO
I wonder how busy the NVIDIA5 will be throughout holiday season. Till Next Time,
Noah on AI
👋 Meet the Hoomans!
Chat with lifelike AI Humans who see, listen, and respond with real emotion. Whether it’s Maxine the witty Gen X’er or The Storyteller with a voice full of mystery, every conversation feels human.
💬 Try it now at Hoomans.chat where your AI talks like one of us.
