🌟 Editor's Note: Recapping the AI landscape from 09/02/25 - 09/08/25.
🎇✅ Welcoming Thoughts
Welcome to the 9th 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.
Back to startups: Exciting interview this week with a Gen Z Founder.
I went to an AI event in Cleveland Monday night. ~200 people in attendance: 80% software engineers.
I’ve thought Perplexity has been overrated for some time now, but I’m impressed with recent deep research improvements.
GPT voice mode has slowly become very good.
Money is not real when it comes to the NVIDIA5.
Interesting news on AI hallucinations in the OpenAI section below.
Crazy week for OpenAI in general.
It’s never been more important to start learning AI, especially if you are, or will be, looking for a job in the next 12-18 months.
Let’s get started—plenty to cover this week.
👑 This Week’s Winner: OpenAI // ChatGPT
OpenAI Acquired Statsig for $1.1B and it’s not even their most impactful move of their week. Statsig is a product testing startup, focused on A/B testing and experimentation. The acquisition will boost OpenAI’s ability to create and iterate on their user-facing apps, namely ChatGPT. Statsig CEO Vijaye Raji will become the new CTO of Applications for OpenAI.
Now for the important stuff: OpenAI announced a $10 billion partnership with Broadcom to design and manufacture its own AI Chips. This is a major move given their current dependence on NVIDIA. Production is expected to start in 2026, with chips tailored specifically for OpenAI’s workloads. This could also mark an industry shift from GPUs (NVIDIA’s specialty) to ASIC chips, which are more efficient but less flexible.
In other big news, OpenAI released a paper detailing why they believe LLM’s hallucinate. They believe evaluation models that guide the AI’s responses are over-encouraging guessing due to the possibility of correctness. The paper states that hallucinations may not be 100% solvable but they can be significantly reduced.
OpenAI also announced:
1. Branching feature to let users spin-off chats but keep the context.
2. Parental controls coming within a month for teen accounts.
3. Crisis routing to better models for conversations deemed high stress.
In addition to all of this it is also worth noting that OpenAI’s projected spend over the next 4 years has jumped from ~ $40B → $120B.
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
Lots of Money: Anthropic closed a $13B funding round at a $183B valuation, with run-rate revenue soaring from $1B → $5B in 2025; Boosted by $500M from Claude Code since its May launch. I’d have invested too if I had $10B lying around.
Copyright Numbers Released: Anthropic agreed to a $1.5 billion settlement with authors over pirated training books ~ $3,000 per book. They also committed to deleting the infringing datasets. Ideally the cleanup doesn’t diminish performance.
Job Displacement: CEO Dario Amodei doubled down on his warning that AI could eliminate up to 50% of entry-level white-collar jobs within five years. Contrary to NVIDIA CEO Jensen Huang’s claims. Baseline AI knowledge will become a common job req within the next 18 months.
🔵 Meta // Meta AI
Political pressure: Senator Edward Markey renewed his demand that Meta bar minors from using its AI chatbots. He accused them of ignoring his 2023 safety concerns.
Meta DARLING: Meta unveiled their Diversity-Aware Reinforcement Learning framework, that trains AI to balance accuracy with diversity in its answers, producing more creative and varied results. Encouraging model for a place where AI often stumbles: creativity.
Meta Expanding Chatbots: Meta is hiring contractors who are fluent in Hindi, Indonesian, Spanish, and Portuguese to design culturally tailored AI personas for those respective markets. Persona chatbots is becoming a core focus for Meta. So far it has caused more harm than good.
⚪️ NVIDIA
Pushback on Regulation: Nvidia is pushing back hard against the GAIN AI Act, a piece of proposed legislation that would force U.S. chipmakers to sell to domestic buyers first. I thought the profit-share solution was a formidable one.
China Still Hungry for Chips: Despite political pressure, Chinese giants like Alibaba, ByteDance, and Tencent are still lining up for Nvidia’s H20 chips, and eyeing the upcoming B30A, showing demand isn’t slowing down.
Cisco Partnership: Nvidia, Cisco, and VAST Data launched the Secure AI Factory, a platform that combines their hardware and software to help enterprises build and deploy secure AI systems more easily.
🟣 Google // Gemini
Antitrust Setback: The U.S. DOJ banned Google from bundling Gemini with Chrome or Search, and ordered it to share key search index and user interaction data with competitors to level the playing field. Could have been way worse for Google (Chrome or Android divestment).
Education Growth: Gemini for Education is now available at 1,000+ U.S. colleges, reaching 10M+ students. The program is aimed at embedding Gemini’s AI into academic workflows and personal learning.
Triple Feature Update: Gemini now supports longer audio uploads, Google Search’s AI Mode expanded to five new languages, and NotebookLM can now auto-generate study guides, quizzes, flashcards.
🔴 xAI // Grok
Tesla and xAI: At Tesla’s November 6 annual meeting, investors will decide on a shareholder-led proposal for Tesla to invest in xAI, targeting synergies in robotics and energy. Been waiting on this since Issue 01.
More Hackers: Security researchers discovered an exploit where attackers trick Grok into embedding malicious links inside promoted posts, letting them bypass platform protections. Commonplace now for xAI. Granted they may be disproportionately open to cyberthreats as a social media platform.
Win in Court: A federal judge granted xAI a temporary order blocking ex-engineer Xuechen Li from joining OpenAI’s generative AI projects, citing the potential trade secret misuse.
🚑 Impact Industries 💤
Medical // AI Created Antibiotics
Penn researchers built AMP-Diffusion, an AI system that generated 50,000 potential antimicrobial peptides and narrowed them to 46 top candidates. In mouse tests, two of the AI-designed peptides cleared infections as well as FDA-approved antibiotics like polymyxin B and levofloxacin, with no side effects.
Research // AI that Dreams
A new framework proposes a wake–dream cycle for AI: learning from data when awake, then “dreaming” through free exploration. In dream mode, models remix past knowledge, imagine “what if” scenarios, and test new ideas. This helps them uncover hidden structures and spark fresh hypotheses beyond standard training.
🎙 Weekly Interview: Andrew Roth
Andrew Roth
🏠 Background: Cleveland native, Andrew earned a B.S from Vanderbilt University, and served as President of the Vanderbilt Innovation and Entrepreneurship Society.
💼 Work: Andrew is the Founder & CEO of dcdx, a global research and strategy firm helping brands like Spotify, Chipotle, Crocs, and L’Oréal attract Gen Z.
🚀 Quote: “AI is accelerating at a pace I’ve never seen before, and how we integrate it is fundamentally different now compared to 12 months ago. There are pros and cons, but it’s not going anywhere”
🎙 Condensed Interview – Andrew Roth, Founder of DCDX & Offline
Noah: Tell me about yourself.
Andrew: I'm Andrew, founder of DCDX, a Gen Z insights and strategy firm, which I've been running for five years. I also recently launched a company called Offline, which helps brands connect with in-person communities. I'm 26, almost 27, and based in Chicago.
Noah: What's your take on the AI space today?
Andrew: It's changing so quickly that you should ask me again tomorrow. Everything is accelerating at a pace I've never seen before, and how we integrate AI is fundamentally different now compared to 12 months ago. There are pros and cons, but it's not going anywhere.
Noah: Where are you using AI in your business workflows?
Andrew: The most common workflow is transcribing and taking notes on calls and interviews. This allows me to be more genuinely engaged in the conversation. I also use AI to create knowledge bases by uploading project scopes or decks, and for generating templates for operational and organizational tasks.
Noah: Why do you prefer using Claude?
Andrew: I prefer using Claude because of its parent company, Anthropic, and their "human first lens" and moral structure. Sam Altman from OpenAI seems to be in a "money race" to be first, and I find it concerning that AGI is being pursued "without limits". I'm in favor of companies that consider the consequences of what they're doing.
Noah: How has AI changed your efficiency?
Andrew: I'd say I have about "one and a half of me now". I can get a lot more done in a day and at a lower cost. For things like legal or financial questions, AI can get me 80-90% of the way there, so I can ask a sharper question that only takes a lawyer 10 minutes of their time.
Noah: What are your thoughts on AI psychosis?
Andrew: It's an important topic, and it's scary because it's real and it's happening. It will likely happen more often given the current loneliness crisis, as AI offers an easy, convenient solution. I wish I had better answers for safeguards, but ultimately, if we prioritize speed and being first, these consequences will happen. It's a problem we have to address because if we can't, we don't deserve to be building the future.
Noah: What advice would you give to a young person entering the workforce?
Andrew: Learn everything you can about AI tools and how to work with them, but don't become dependent on them. Dependency is what AI will take over. You need to maintain your independence in thought, because if your thinking follows the AI's, everything will blend together and you won't produce novel ideas. The difference won't be in how they use it, but "actually how they don't".
👨💻 Practical Use Case: GPT Projects
Difficulty: Beginner
Most people use ChatGPT in a one-off way: open a chat, ask a question, close the window. But GPT Projects take things further by giving the model memory and context across an entire workspace.
With Projects, you can:
Store project wide instructions: Instead of pasting the same context over and over, you can set persistent instructions the model always follows.
Keep better memory: The model remembers what you’ve uploaded, what you’re working on, and the style you prefer, so you don’t have to repeat yourself.
Organize files and tasks together: Drop in documents, notes, or datasets, and the model can reference all of them in one place.
Example: Imagine you’re writing a research paper. In a normal chat, you’d have to re-upload your sources every time and restate your formatting rules. In a GPT Project, you upload your sources once, set your style guide in the instructions, and the model can consistently help you draft sections, generate citations, and keep the tone aligned across the whole paper.
I’ve found projects useful in not only organizing my chats, but also the context aspect referenced above. It still leaves something to be desired but definitely a step up from scattered chats. Create a project folder and try it out if you haven’t already!
🏞️ Startup Spotlight
Offline
Offline: Where online communities end, Offline begins.
The Problem: Gen Z is gathering. Not on Facebook or Discord, but in clubs, pop-ups, game nights, and dinner parties. These grassroots in-person communities have exploded post-COVID, but most can’t monetize. They’re left without infrastructure, sponsorships, or scalable revenue.
The Solution: Offline helps these IRL communities thrive by bridging them with brand partnerships. By building the rails for local monetization, Offline is letting Gen Z creators earn by doing what they’re already doing: bringing people together.
The Backstory: Offline spun out of dcdx, Andrew Roth’s Gen Z research firm behind insights for Chipotle, Spotify, Crocs, and more. After years tracking culture and community, Roth saw the shift firsthand: digital fatigue, loneliness, and a hunger for offline connection. With early traction from brands and a mission to empower real-world belonging, Offline is shaping the future of community monetization.
Contact: [email protected]
“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 I noticed people checking their phones during the AI event speaker sessions, it wasn’t texts or browsing, it was GPT and Claude. Till next time,