🌟 Editor's Note: Recapping the AI landscape from 09/23/25 - 09/29/25.
🎇✅ Welcoming Thoughts
Welcome to the 12th edition of NoahonAI.
What’s included: company moves, a weekly winner, AI industry impacts, practical use cases, and more.
AI Picture is really shaping up: OpenAI/Chat are dominating the consumer space, and Claude has a strong lead in the enterprise arena.
Insightful interview this week with a good friend of mine on Robotics, Computer Vision, AI Training, and Machine Learning.
Strong week across the board for the NVIDIA5.
The vibe coding technology is only getting better, researching the second impact industry down below blew me away.
I think we’re gonna see a new term/branch for humanoid robotics soon. Differentiating between a home ‘butler’ robot, and a factory ‘worker’ robot.
Very exciting to see these companies dive into robotics.
I think we’re gonna miss the good ole days where there were no ads when we messaged LLM’s.
If anyone is in Ecommerce reading this, practice your GEO!
I’ll be speaking at an AI panel in Columbus on Saturday (10/04) and leading two AI workshops. Check it out if you’re in the area!
Let’s get started—plenty to cover this week.
👑 This Week’s Winner: OpenAI // ChatGPT
Another Strong Week for OpenAI. They started the week off with the launch of GPT Pulse, a new daily briefing feature for Pro users. Pulse proactively does overnight research, drawing from chat history, memory, and connected apps, and surfaces personalized “cards” each morning with updates on tasks, interests, or plans.
While Pulse is exciting, the bigger news came Monday afternoon with the announcement that ChatGPT E-commerce was officially here. With the help of financial platform Stripe, GPT users can now make in-chat purchases from Etsy and Shopify. Other integrations are said to be coming soon.
It’s also worth noting that this may be the opening OpenAI was looking for in a potential move to in-chat ads. The speculation comes from OpenAI’s search for a ‘Head of Ads’, but we don’t know what it will look like quite yet. I hope the ads are not over the top when they do inevitably arrive.
In other news:Marketing push: OpenAI rolled out a major TV campaign showing ChatGPT in everyday use cases like cooking, learning, and planning.
Enterprise growth: The company has quietly expanded its GPT Business functionality, adding new connectors for Gmail, Calendar, Teams, GitHub, Dropbox, and more.
Infrastructure buildout: OpenAI, alongside Oracle and SoftBank, announced five new U.S. data center sites under its $500B “Stargate” expansion.
Next-gen video: OpenAI is preparing to launch Sora 2, it’s latest video model, along with a TikTok-style app for AI content, powered by said model.
-
Really strong week for OpenAI. Intrigued to see how E-commerce may change over time based on in-chat buying. Stargate seems to be underway which is game changing in its own right. The only move I don’t like is the AI-slop Tiktok clone. Not sure what the benefit is there.
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.
🟠 Anthropic // Claude
Released Claude Sonnet 4.5: Anthropic is calling it a breakthrough in coding and agentic AI. The model outperformed prior versions on SWE-bench, sustained 30-hour autonomous coding runs, and showed stronger “computer use” skills for building real applications. Good stuff.
Claude Comes to Copilot: Sonnet 4.5 is now integrated into Microsoft 365 Copilot and Copilot Studio, giving enterprises Claude as an alternative to OpenAI’s models. Interesting to see Microsoft opening its doors given their stake in OpenAI.
Anthropic is Growing: Plans to 3x its international workforce and 5x its applied AI team, citing rising global demand. Appointed Chris Ciauri, a former Google Cloud and Salesforce executive, as Managing Director of International to lead the expansion.
⚪️ NVIDIA
NVIDIA Moves Further into Robotics: Partnering with Google DeepMind and Disney Research, released Newton, an open-source physics engine for more realistic robot training. Deepmind + NVIDIA collab is awesome. Lots of smart people and AI models in that room.
Humanoid Focus? Along with Newton, NVIDIA unveiled Isaac GR00T N1.6, a humanoid reasoning model that combines vision, action, and language to help robots plan and execute multi-step tasks. Very interested to see how humanoid robotics evolves over the next few years.
Stock Surge Following OpenAI Deal: NVIDIA stock hit a record high, jumping 4%, after announcing the landmark partnership with OpenAI last week.
🟣 Google // Gemini
Launched Gemini Robotics 1.5: Built by Deepmind (Google’s AI research lab). Release comes with embodied reasoning (understanding surroundings, planning, and using tools), VLA (understanding → motor skills), and motion transfer (skills learned on one robot transferred to others). Cool stuff here.
Gemini Mobile Gaming: Introduced Gemini Live for games on the Google play store. An in-game AI assistant that provides real-time tips without leaving the screen. ‘Gaming’ will be a big notch in the AI race belt for whoever gets there first on the frontend. NVIDIA already there on backend.
Gemini Gems: Users can now share their custom AI “Gems” (specialized mini agents within the gemini system) via links. Don’t hear much buzz about these. Similar to custom GPT’s in Chat.
🔵 Meta // Meta AI
Meta Launches Vibes: A new feed of short AI-generated videos in the Meta AI app. Users can create or remix AI clips and cross-post to Instagram or Facebook. Critics calling much of the content “AI slop”. Critics are right. The worst use case for AI on display here.
Meta Dating: Meta added an AI assistant to Facebook Dating that helps refine profiles and find matches through natural prompts. A new “Meet Cute” feature also delivers a weekly surprise match, now rolling out in the U.S. and Canada. Wonder if they’ll try and go full Black Mirror.
Meta U.S. Defense: Meta is extending its open-source Llama model into national security, powering tools for U.S. Special Operations and making it available to agencies, contractors, and allies.
🔴 xAI // Grok
xAI for Gov’t Begins: The contract covers Grok 4 and Grok 4 Fast, which xAI describes as its most advanced reasoning models. The agency said xAI engineers would also assist agencies with implementation. Charging $0.42 per year so they can say its the cheapest - Others charging $1 year.
Another Lawsuit: xAI filed a lawsuit accusing OpenAI and former employees of stealing proprietary code, business plans, and strategies. As previously reported, one engineer was singled out for allegedly taking files. I’d love xAI to stay focused on what they’re great at and not constant distractions.
Will Grok 5 Reach AGI? Elon said he now believes xAI’s upcoming Grok 5 could reach AGI, citing strong benchmark results from Grok 4. Bold claim, but would propel xAI towards the top.
💻 Impact Industries 🤖
Enterprise SaaS // Cloud Apps
Lovable, a U.S. startup, has introduced new cloud and AI features that let developers build and deploy full-stack applications with minimal code. The platform uses AI to create app components, manage databases, and set up hosting directly in the cloud. This removes much of the friction in going from idea to production. They had a cool demo this week, giving a 10yr old a PC and letting him build a complete application, full stack with API's and all, just from some text. It's not perfect yet, but a long ways away from just a couple years ago.
Robotics // Industrial Automation
Sereact has launched a reality-trained AI robotic system designed for industrial deployments, capable of adapting to new tasks without hand-coding. The platform uses a vision-language-action model trained in simulated and real environments, enabling robots to understand written or spoken commands and execute them safely in warehouses or factories. By learning from both synthetic and physical data, Sereact aims to bring more flexible, general-purpose automation to industries where rigid, pre-programmed robots still dominate.
🎙 Weekly Interview: Joey Houser
Joey Houser
🏠 Background: Joey is from Cleveland, OH. He earned a B.S. in Math & CS at Fordham, and is finishing his Master’s in CS at CWRU. He previously worked as an AVP at Citibank in Residential Real Estate.
💼 Work: Project Integration Engineer at ReadySet Surgical, with research spanning industrial crystal growth, computational biology, and macroeconomic modeling.
🚀 Quote: “I really see the connection of AI into the physical world with robotics and computer vision as a huge segment of growth going forward”
📄 Condensed Interview Transcript
Noah: Where do you see the next major wave of growth in the AI space?
Joey Houser: While LLMs are powerful, I don’t see massive future growth there beyond greater accuracy. The biggest advancement will come from connecting AI to the physical world—especially robotics and computer vision.
Noah: What is the significance of “testing drugs in silico”?
Joey Houser: It means simulating a drug’s interaction within the human body on a computer before clinical trials. This could dramatically lower costs, improve predictions, and accelerate drug creation.
Noah: Do you see a future for humanoid robots in the near term?
Joey Houser: I’m skeptical of universal humanoid robots—they’re mostly a marketing gimmick. The real power lies in specialized robots built for specific, dangerous, or repetitive tasks, like robotic arms for welding or unloading trucks.
Noah: What is Tesla’s core advantage in the AI/robotics race?
Joey Houser: Tesla has one of the world’s largest labeled visual data sets, collected from every car they’ve built. This is an invaluable, high-cost asset for training self-driving and robotics models.
Noah: What career advice do you have for someone starting in AI or CS?
Joey Houser: Read a lot—staying informed is hugely valuable. And network in non-traditional ways: talk to friends’ parents, older people in bars, anyone who can give you diverse perspectives and open career paths.
👨💻 Practical Use Case: Explore GPT’s
Difficulty: Basic
If you’re in ChatGPT, you may have seen the Explore GPTs tab in the top left. This is where you’ll find specialized versions of ChatGPT. We’ve talked in the past about building a GPT Wrapper, and this is exactly that, except its sits in ChatGPT’s version of the app store. A custom GPT comes preloaded with rules, knowledge, and even connections to outside tools.
Some of the most popular ones right now include:
Astrology Birth Chart GPT – answers questions based on your birth info.
Scholar GPT – pulls from PubMed, arXiv, and other academic sources.
Fitness, Workout & Diet Coach – gives personalized health insights using a large dataset.
Humanize AI – rewrites AI text to sound more natural.
Image Generator – creates and refines visuals in different styles and tones.
Technically, a Custom GPT is just ChatGPT with a system prompt (hidden instructions) plus any files, APIs, or knowledge sources it’s connected to. That context loads automatically every time you open it, so it feels like a mini-app designed for one job.
You don’t need to build one to benefit. Most people just browse, pick one, and start using it. But if you want your own, go to the explore section in ChatGPT and click create in the top right. Or, if you’d like one for your website check out this blog.
🎥 Startup Spotlight
Creatomate
Creatomate: Automate video like it’s code.
The Problem: Creating dynamic video content at scale is a pain, especially for developers, and SaaS platforms who need to personalize thousands of videos without expensive editing tools or massive production teams.
The Solution: Creatomate is a no-code & API-first platform that turns templates into fully-automated video production pipelines. Users can generate personalized videos from data, dynamically swap images/text/audio, and create social-ready visuals without lifting a finger. For developers, it functions like a programmable video engine.
The Backstory: Founded in 2020, Creatomate was built for teams tired of making the same promo videos manually. With a growing customer base across marketing, HR, and SaaS industries, Creatomate is now powering everything from real estate listings and onboarding flows to AI-generated video campaigns; quietly becoming the go-to infrastructure for automated video creation.
My Thoughts: A tool I started using a few weeks ago for automating video content. Still early but like it a lot so far. There’s a ton of AI video making platforms out there, but I haven’t found many besides Creatomate with flexible API’s for when you already have a content idea in mind.
“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 what we’ll be able to build from just a few lines of text in 10, 20, or 30 years. Till Next Time,
Noah on AI