In partnership with
🌟 Editor's Note: Recapping the AI landscape from 08/26/25 - 09/01/25.
🎇✅ Welcoming Thoughts
Welcome to the 8th 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 the clearest picture of the AI space.
Great interview this week with a former professor of mine on AI in education.
Lots of legal disputes in the news as of late. May be interviewing a law professor soon.
Jevon’s paradox is becoming increasingly accurate in AI consumption.
Definitely some concern with how LLM’s are handling mental health.
Working on getting more full interviews up on Youtube.
Expanding Startup Spotlight to companies outside my network.
Finding myself increasingly bullish on Google and Anthropic.
Finding myself increasingly bearish on Meta.
Let’s get started—plenty to cover this week.
👑 This Week’s Winner: NVIDIA
Not a runaway victory from NVIDIA this week, but some good numbers nonetheless. Earnings week, enterprise adoption, and a strong forecast for Q3.
NVIDIA’s Q2 earnings hit a record $46.7 billion, with $41.1B driven by data center sales. In addition to revenue from the companies listed below, NVIDIA is seeing strong adoption in the enterprise market. Companies like Disney, Foxconn, and TSMC are rapidly adopting RTX PRO servers to power their AI workloads without reworking their entire infrastructure.
One concern? Revenue concentration. Nearly 40% of NVIDIA’s Q2 revenue came from just two customers, raising questions about long-term stability. However, as long as AI continues to be dominant, I don’t see NVIDIA’s numbers going down. The company is projected to do $54 billion in Q3.
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.
🟣 Google // Gemini
Nano Banana is Exploding: Google’s new image model, also known as Gemini 2.5 Flash or Nano Banana, is getting rave reviews. It allows for precise image edits via simple chat, while preserving faces and detail. Google may now have the best image editor AND video generator.
Hey Siri, Please Improve: Apple is reportedly in early discussions with Google to leverage Gemini AI to power an upgraded version of its Siri voice assistant. Apple has been absent from AI, and Siri is in a bad state. This would be a welcome move for both companies.
Accused of Breaking Safety Pledge: 60 British lawmakers accused Google DeepMind of violating its Frontier AI commitments by releasing Gemini 2.5 Pro without timely safety data. Seems their concern is more-so on the precedent, not the product itself.
🟠 Anthropic // Claude
Claude for Chrome: Anthropic released a browser extension that lets a Claude AI agent live inside your Chrome sidebar. It can see your current webpage, chat with you, and even perform inline tasks. Access is limited to only 1,000 Max users for now. Very cool. Browser wars are just beginning.
Historic Copyright Settlement: A group of U.S. authors have reached a class-action settlement with Anthropic in a case involving pirated books used to train the AI model. Purchased books were ruled fair game for training in June.
New Data Policy: Users must opt out by September 28th if they don’t want their Claude chats and code included in model training. User data will be legally trained on for up to 5 years; Up from 30 days. Does not include business, govt., education, or API users. No real problem with this.
🔴 xAI // Grok
Launched CodeFast‑1: A high-speed, agentic coding model, to compete with Claude Code and others. Available for a limited time with select partners, including GitHub Copilot. Excited to try it out.
Grok for Government is Back: After recent exclusion, Grok 3 and 4 were added back to the GSA federal vendor list due to White House intervention.
Lawsuit over Trade Secrets: xAI sued a former employee for allegedly stealing trade secrets and bringing them to OpenAI after he was hired by Sam Altman and co. last month.
🟢 OpenAI // ChatGPT
Lawsuit after Teen Suicide: A California family is suing OpenAI and CEO Sam Altman after their 16‑year‑old son, Adam Raine, died by suicide. The lawsuit says he died using the exact method GPT described and validated. Obviously an issue. Need better AI safeguards when it comes to mental health.
How is OpenAI Responding? GPT is exploring better crisis intervention options, including connections to certified therapists, parental controls, and introducing emergency contacts.
API Voice: OpenAI just released an advanced API voice model. It will allow developers a much easier time when building voice agents in third-party apps. Can make calls and chat cohesively in real time. This is huge!
🔵 Meta // Meta AI
Meta to Release New Model by Year’s End: Amidst the organizational shakeup and other issues, Meta is looking to a fresh start by attempting to release their new Llama 4.5 model by the end of 2025.
Talent Exodus and Scandal: Meta lost 3 key hires who decided to return to OpenAI after less than a month with the company. Additionally, Meta was found to have unauthorized celebrity chatbots making sexual advances to users. The week where “flirty chatbots gone wrong” is out of the news will be a good week.
Meta Back in Politics: Meta is launching a California-focused super PAC, to support political candidates who advocate for lighter AI regulation, ahead of the 2026 CA governor’s race.
🚑 Impact Industries 📚
Medical // Stroke Care
England’s NHS has deployed an AI tool across all 107 stroke centers that analyzes CT scans in under a minute, slashing diagnosis and treatment times from 140 minutes to 79. In nationwide trials, the faster workflow tripled recovery rates, with 48% of patients regaining independence versus 16% before. The rollout will reach 80,000 patients a year, marking one of the largest and most impactful deployments of AI in frontline medicine.
Education // AI in School
Bellingham Public Schools in Washington are introducing structured AI use into classrooms this fall, allowing students to use AI under set rules. Students must fact-check outputs and disclose usage, while teachers oversee responsible integration. The policy aims to build AI literacy without encouraging shortcuts, treating AI less as a forbidden tool and more as a skill students need to master ethically for the future.
🎙 Weekly Interview: Professor John Kerezy
Professor Kerezy
🏠 Background: Long-time educator and communicator, with a B.A. from Wabash College, an M.A. from The Ohio State University, and an Ed.S from Kent State University.
💼 Work: Faculty member at Tri-C since 2004. Former journalist. PR practitioner, Author, and founder of Kerezy Communications.
🚀 Quote: “Too many professors and too many high school teachers are feeling like they're police officers. It's a bad path for everybody. You gotta somehow figure out a way to overcome that.”
🎙 Condensed Interview Transcript
Noah: What are the ramifications of having AI in the classroom?
Prof. Kerezy: The risk is overreliance. If students lean on AI too much, it weakens their ability to develop critical thinking. I’d ban it in lower elementary, introduce it gradually in upper grades, and set course-specific rules in high school and college. The challenge is finding balance between AI as a shortcut and AI as a spark for curiosity.
Noah: Where do you see AI being most effective in teaching?
Prof. Kerezy: When it assists teachers with lesson prep, freeing up time for instruction, or when it helps students outline ideas and explore pro/con debates. If used to enhance critical thinking, it’s powerful. If used just to finish tasks faster, it harms learning.
Noah: How do we prevent cheating while allowing AI for positive use?
Prof. Kerezy: It’s tough. Some schools are banning phones, and some professors are going back to in-class handwritten exams. At the end of the day, you can’t stop students from using AI at home. The solution is teaching what’s good and responsible, guiding them toward brain development instead of shortcuts.
Noah: Who should shape AI adoption in education?
Prof. Kerezy: Faculty must take the lead, with administrators there as well. If not, tech companies will set the rules based on the dollar not the student. Community colleges especially should step up to teach about AI and prepare the public, rather than letting business interests dictate how it’s used.
Noah: Any final thoughts for where we are with AI?
Prof. Kerezy: Think of the first digital CDs: recorded, mixed, and output digitally. With AI, the inputs, processes, and outputs are also changing. The key is whether we’re mixing human thought with AI or letting AI dominate completely. That’s the balance we need to strike.
👨💻 Practical Use Case: GEO (Generative Engine Optimization)
Difficulty: Mid-level → Advanced
To explain GEO we need to start with SEO. SEO, or Search Engine Optimization, is probably familiar to most website builders or anyone in marketing, advertising, or copywriting. SEO is essentially the practice of making your website, blog, service, or product appear more often in organic searches. So, when someone types “best dentist in Cleveland, OH” into Google, your website pops up first. The way this is achieved is through a combination of domain authority (making your site trustworthy), keywords, and backlinks (other sites linking to you).
So how is GEO related? GEO, or Generative Engine Optimization, is essentially the same practice, but instead of trying to rank on the first page of Google, you are attempting to be a recommended business by ChatGPT, Gemini, or any LLM.
Biggest differentiation: With SEO, you optimize for an algorithm that ranks websites. With GEO, you optimize for an LLM that answers directly, so the focus shifts from keywords, to being a reliable, structured source the model can trust.
Starting focus:
Make your info easy to read: Keep your business details (address, hours, services, pricing) up to date and in clear formats online. Models need clean, structured data they can pull from.
Get listed in trusted places: Directories, review sites, and local organizations are where LLMs often look first. The more your business shows up in reliable sources, the better.
Encourage detailed reviews: LLMs learn from natural, conversational reviews. Instead of short “Great service!” reviews, encourage customers to explain what they liked in full sentences. It’s easier for models to surface and recommend.
I have a friend who’s been an SEO expert as long as I’ve known him. He’s big on GEO, and I’d recommend giving him a follow on Linkedin if you want to stay up to date with best practices. Good luck!
📲 Startup Spotlight
Reacher App
Reacher: An AI‑powered platform to automate and scale creator marketing.
The Problem: In a world where short-form content is key, creators have never been more valuable to marketing teams, but it’s hard to find the right ones. Teams often spend countless hours discovering creators, sending outreach messages, and tracking campaigns across TikTok and other platforms.
The Solution: Reacher uses AI to automate the end-end creator marketing process. From discovery, to outreach, to pipeline management, campaign execution and payment, Reacher handles it all.
The Backstory: Reacher was founded in 2024 by Bora Mutluoglu (a Bay Area native who worked at Palo Alto Networks) and Jerry Qian (formerly at Meta and NASA). After seeing first‑hand how time‑consuming creator outreach and campaign management can be, they built Reacher as an “AI Brand Manager” for e-commerce businesses scaling through the creator economy. The startup participated in Y Combinator’s Summer 2025 batch.
Contact: Interested brands can book a demo here.
“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
When someone asks you a question that you don’t have the answer to, do you say “Google it” or “Ask ChatGPT”? Till next time,
Noah on AI
The easiest way to hire globally
Don’t let borders limit your hiring strategy. RemoFirst gives you one platform to legally employ talent around the world—compliantly, affordably, and fast.
We support EOR and contractor payments in 185+ countries with no annual contracts, flat pricing, and full transparency.
Whether you’re a startup or scaling enterprise, you’ll get hands-on support and built-in tools for international payroll, health benefits, taxes, and more.
RemoFirst offers a free tier for contractor management, and EOR fees start at $199/month.