In partnership with
🌟 Editor's Note: Recapping the AI landscape from 09/30/25 - 10/06/25.
🎇✅ Welcoming Thoughts
Welcome to the 13th edition of NoahonAI.
What’s included: company moves, a weekly winner, AI industry impacts, practical use cases, and more.
Welcome new subscribers, I’d recommend checking out Issue 01, and Basics & Buzzwords for an idea of how we got here.
Excited to announce a new title sponsor at NoahonAI. Craig is a friend of mine and he’s building a great company. Highly recommend checking out VERN AI.
Had a great time speaking at the NOW Foundation event this past weekend.
In one of the workshops I led there was someone who had only used GPT just once, and someone who had already automated part of their business. Very cool to see the range of AI adoption!
Was pretty unimpressed with the launch of GPT-5, since then i’ve been very impressed by the speed and moves of OpenAI.
Throwback interview this week and a good one lined up for next week.
I started building out a prompt library (a list of AI prompts for different tasks), I’d recommend doing the same.
Everywhere I go I hear more comps of the AI boom to the Internet Boom.
Some of these SORA videos are insane - see below.
It’s not generally the question for a company that just raised $100B but… does OpenAI have enough capital?
For the first time in a while it was a pretty slow AI week, outside of OpenAI.
It’s rumored that Apple CEO Tim Cook may be exiting soon, could this mean Apple may FINALLY enter the AI race… definitely a story to follow.
Let’s get started—plenty to cover this week.
👑 This Week’s Winner: OpenAI // ChatGPT
OpenAI continues to roll. This week, similar to last, is headlined by a major chip partnership. OpenAI signed a multi-billion dollar agreement with AMD to access up to 6 gigawatts of power via their GPU’s.
This is a big agreement coming off their NVIDIA deal, and is said to be more about the industry wide demand for compute, rather than a shift away from NVIDIA. Whether intended or not, the diversification is not a bad idea.
In other news, OpenAI had its dev day on Monday and it came with some interesting news, and some potential game-changing products:Sam Altman announced that ChatGPT has now reached 800M weekly users. Nearly one of ten people on the planet are using the tool weekly. Astounding.
OpenAI is launching a tool that lets users chat with apps. developers can build apps that run inside ChatGPT, and users can talk to them. First apps include Canva, Zillow, Spotify, Figma, and more. This is a huge deal!
In addition to app building, GPT is launching AgentKit to help developers build workflows within the GPT system. It’s similar to n8n which we’ll discuss later.
Last but not least, GPT released Sora 2 this past week, and it is incredibly impressive. While the app itself is essentially AI Tiktok, the technology demonstrated rivals the top AI video tools. The app lets users create and post AI created video, and remix video with the AI avatars of other people.
It is the most impressed I’ve been of an AI launch since Gemini’s Veo3, and it may be even more realistic. It is also very damning, as we will soon be at a point where AI created video and recorded video are indistinguishable.
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
Deloitte Partnership: Anthropic partnered with Deloitte to deploy Claude across 470k employees, its largest enterprise rollout yet. A new Claude cert program will train 15,000 Deloitte staff to scale safe AI in regulated industries. Another win for enterprise Claude.
Are You Testing Me? Anthropic’s newest model, Claude Sonnet 4.5, displayed evolutionary awareness in about 13% of experiments, detecting when it was being tested. Likely just recognized testing patterns, but could hint at meta cognition → AGI/ASI.
Slack Integration Launch: Claude can now work directly inside Slack (workplace messenger), summarizing threads, prepping users for meetings, and drafting responses. Good move. Especially with the OpenAI news.
⚪️ NVIDIA
Fujitsu Partnership Expanded: NVIDIA and Fujitsu are teaming up to build end-to-end AI infrastructure. The collaboration will power AI agents for healthcare, robotics, and manufacturing, starting in Japan and expanding globally. Cool.
Enfabrica Acquisition: NVIDIA acquired Enfabrica, a chip startup specializing linking huge GPU clusters, for $900M. The move strengthens NVIDIA’s architecture and helps reduce bottlenecks in large-scale AI systems.
China Warning: CEO Jensen Huang warned that U.S. chip export bans could backfire by pushing Chinese companies to catch up to survive. He said China is now “just a few nanoseconds behind”. This has been a tricky line to walk since the start. Not sure there’s a right answer.
🔴 xAI // Grok
Grokipedia Incoming: Elon confirmed that xAI will roll out Grokipedia v0.1 in two weeks: an AI-powered encyclopedia designed to “massively improve” on Wikipedia by spotting bias, missing context, and factual errors.
AI Gaming: xAI is recruiting video-game developers and “game tutors” ($45–$100/hr) to teach Grok about gameplay mechanics, storytelling, and design. Plans to release a fully AI created game by the end of 2026. Very intrigued by the future of AI gaming, especially with AR/VR. Smart move.
Colossus Expansion: xAI plans to spend $18B on 300K new NVIDIA chips for its Memphis Colossus data center, aiming to reach 1M chips and become the world’s first gigawatt-scale AI supercluster. Billions and billions going towards compute right now.
🔵 Meta // Meta AI
AI-Driven Ads: Starting Dec. 16, Meta will use Meta AI interactions (text and voice) to refine ad targeting on Facebook and Instagram. The change applies only to users who engage with Meta AI. Makes sense, but I use Meta more for advertising than AI.
Rivos Acquisition: Meta plans to acquire Rivos, a chip startup valued near $2B, to reduce reliance on NVIDIA. The deal reflects Meta’s broader push to internalize AI hardware development. Something of a pattern here with NVIDIA5 companies moving chip-making internal.
Gamified AI Adoption: Meta is tracking their own employee AI usage through dashboards and a chatbot called Metamate, which awards badges and milestones to encourage daily AI use across teams.
🟣 Google // Gemini
Gemini At Your Door: Google unveiled new Nest Cam, a 2K Doorbell, and a Gemini-native smart speaker, all powered by Gemini for Home. Include AI context to identify happenings in the home and at the door. Their 1st move into AI native hardware.
CodeMender Agent: DeepMind launched CodeMender, an AI agent that patches coding vulnerabilities in real time. Uses self-validation to secure massive codebases. Cool stuff. IMO there’s a business opportunity to simply review, fix, and validate code written by ‘vibe coders’.
Gemini Pro Outage: On Sept. 29, Gemini Pro suffered a widespread outage causing errors and slow responses. Google restored service within hours.
🎬 Impact Industries 📲
Creative // Film & Media
Tilly Norwood, an AI-generated actress created by director Justin Kurzel and his team, is blurring Hollywood’s boundaries between performance and simulation. Fully synthetic yet photorealistic, Tilly can act, emote, and even age across films without ever stepping on set. The project has received mixed reviews in Hollywood, as studios weigh the promise of limitless production against fears over the future of human actors. Guessing this will become a lot more popular in the industry over time. It won't be the top actors who get replaced, rather the little-known. Cool, but morally questionable.
Consumer Tech // AI Companions
Friend, an AI companion startup, drew viral attention after its New York subway ads were defaced by commuters labeling it “creepy” and “surveillance tech.” The company, which lets users create personalized digital friends through voice and text, called the graffiti “entertaining” and part of the cultural conversation. The backlash underscores a widening gap between public skepticism and the rapid mainstreaming of emotional AI in everyday life. Not a huge fan of this. The hardware 'friend' pendant is just plastic, and I don't love the idea of AI friends, but we'll see where it goes.
🎙 Interview Callback → Week 4: Manny Larcher
Manny Larcher
🏠 Background: Based in Columbus, OH, Manny exited a venture-backed startup in 2017. More recently, he was a Venture Analyst, advising early-stage startups across tech and healthcare.
💼 Work: Manny is CEO of Colaeb for business growth and Co-Founder of Stopwatch Creative. He’s a startup advisor, mentor, consultant, investor, and writer.
🚀 Quote: “You should just execute and not overthink. If there was ever a time to not overthink, this might be it.”
📄 Condensed Interview Transcript – Manny Larcher
Noah Weisblat: Manny, tell us about yourself.
Manny Larcher: I’ve spent over a decade at the intersection of tech, marketing, and venture capital. I’ve built startups, advised founders, and helped launch Collaeb — a platform connecting businesses to new customers, capital, and talent. I’m passionate about helping founders execute faster and smarter with AI.
Noah: What’s your general take on the AI space right now?
Manny: It’s one of the most exciting times I’ve ever seen. I use AI daily for efficiency and creativity. Tools like ChatGPT and Taja.ai have changed how I run operations — saving hours each week while improving content and customer engagement.
Noah: How do you use AI at Collaeb?
Manny: I upload long-form videos into Taja.ai, and it creates short-form clips for social posts in minutes. That one workflow boosted our engagement, followers, and even new customers. AI gives small teams big-company output.
Noah: What advice do you give founders trying to stand out in the AI era?
Manny: Just start. Too many people wait for perfect. If you build something valuable and someone pays you for it, that’s success. Execution beats overthinking every time.
Noah: Some people say AI is a bubble — what’s your take?
Manny: Totally disagree. AI is democratizing entrepreneurship. Before, 9 out of 10 small businesses failed; now founders can run leaner and smarter. The real divide will be between those who execute and those who just talk.
Noah: Where do you see AI in the next few years?
Manny: Things are moving too fast for 5-year predictions. What used to take years now happens in months. I focus on what I can do in the next 6–12 months — that’s where opportunity lives.
👨💻 Practical Use Case: Custom Workflows
Difficulty: Advanced
A few months ago we talked about building basic AI workflows with Zapier, now time to take the next step. There’s 3 major no-code → medium code workflow tools that I trust for SMB’s. There’s a few others that are valuable, but they’re a bit more geared to larger enterprise companies.
When I first started building workflows, I looked at these options and decided to try and build something just as usable from scratch with more of a priority on a chat interface. It took a lot of time and effort but I got to a place where I could start uploading JSON blocks (workflow guides) to run within the system.
Here’s what it looked like:
After building the baseline application I had to decide if I wanted to continue to build, use it for my consulting, and potentially turn it into a consumer product OR pivot to using an established tool.
Neither one was a bad idea, but given the heavy lift of integrating the workflow blocks with other apps at scale [Google OAuth certification (setting up sign in with google) is no joke], I decided I’d be better off working with an established tool, and maybe coming back to this at some point.
That brings us to today. The tool I’m currently using is n8n. Think of it like a flow chart of applications that you can turn into meaningful workflows. It’s slightly more complicated than the other two tools, Zapier and Make, but it’s definitely learnable with minimal coding experience.
With n8n, you can build any type of AI workflow and use code/external API’s/Agent nodes, to turn them into super advanced systems. We’ll focus on lighter workflows and just getting an idea of what AI can do when strung together.
OpenAI just released a similar tool which I’m excited to try out as well. Check out this video to get an idea of what n8n is all about!
⏈ Startup Spotlight
Alter Ego
AlterEgo: AI that thinks out loud, so you don’t have to.
The Problem: We’re constantly switching between thoughts and tech, typing, swiping, searching, but what if you could control devices just by intending to speak? For people with motor impairments or those in high-focus environments, seamless, silent interaction is the dream.
The Solution: AlterEgo is a wearable, silent speech interface that uses neuromuscular signals → captured through sensors on your face and jaw, to detect internal speech. That means when you silently say words in your head, with a bit of facial movement, AlterEgo can recognize and translate them into text or commands. It’s not telepathy or mind-reading, rather it’s more like whispering without sound, using EMG (electromyography) to pick up micro-movements and map them to language.
The Backstory: Born out of the MIT Media Lab, AlterEgo has been in the works since 2020. The device straps comfortably to your jawline and syncs with a real-time language model trained on each user’s subvocal patterns. While the interface looks futuristic, the real innovation is giving people a way to interact with machines: silently, hands-free, and without breaking focus. It has massive potential across accessibility, defense, productivity, and immersive computing.
My Thoughts: When I first came across this I was blown away… and a little scared, because I though telekinesis had been achieved. However, we’re still some ways away from that, but this is a very interesting step towards it. Will definitely be following along.
“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 started using 3 side-by-side GPT browsers to help fact check and aggregate data in the race section. Super efficient and lines up pretty well.
Till Next Time,
Noah on AI
VERN Makes AI More Human.
VERN is an Emotion Recognition System (ERS) that detects real human emotions—like anger, fear, joy, and sadness—in every sentence of a conversation. It’s what makes AI truly empathetic, responsive, and trusted.
Here’s how you can use It:
