🌟 Editor's Note: Recapping the AI landscape from 08/04/25 - 08/11/25.
🎇✅ Welcoming Thoughts
Welcome to the 4th edition of NoahonAI.
What’s included: company moves, a weekly winner, AI industry impacts, practical use cases, and more.
Wow Claude Code is Incredible. I finally tried it (after previously using a mix of the Claude LLM and Cursor) and I can’t stop using it.
Spent the weekend building an immersive memory app, similar to what I mentioned in the Week 1 Impact Industry.
Someone asked what programming language I was writing the app in and I realized I never even thought about it. Claude Freaking Code.
OpenAI released GPT-5 and I was disappointed with the result. Read my Blog from release day. The web app is borderline unusable right now.
People are reacting to new AI drops (myself included sometimes) like they’re either going to change the world tomorrow or AI is just a bubble. It’s somewhere in between but closer to the former.
Somewhat concerning: ChatGPT only had positive things to say about its launch, unlike every other LLM.
I’m switching to Claude for a few tasks until GPT gets its act together. Still keeping GPT as my primary.
No interview this week so I tried something kinda wacky instead. Bunch more regular interviews lined up ahead. You’ll see what I mean.
Seems like every week a new Black Mirror concept is coming to life with AI. See today’s impact industry // S6 E1 Joan is Awful.
Let’s get started—plenty to cover this week.
👑 This Week’s Winner: Anthropic // Claude
Another week at the top for Anthropic. If conversation length limits weren’t so strict, I would consider a move to the Claude LLM full time, in addition to Claude Code. The strong week for the company was highlighted by a quiet release of Claude Opus 4.1, global expansion, hacker dominance, and and a move into government.
Opus 4.1 included a 2% increase on SWE benchmarks → up to 74.5% coding accuracy, a refinement on complex debugging skills, and an improvement on agentic tasks alongside real-world coding. The Anthropic team introduced Hidetoshi Tojo as the Head of Japan Operations and it plans to hire hundreds more at its Tokyo office opening this fall. At Carnegie Mellon’s PicoCTF, a Cybersecurity competition put on by the University, Claude outperformed human hackers by rapidly solving 11 of 20 coding challenges and earning 4th place overall. Last but not least, Claude joined OpenAI and Gemini in striking a deal with the U.S. GSA to offer Claude to federal agencies at $1/year.
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
Focus on Students: Google is offering free AI Pro subscriptions to students for a year. They also just launched Guided Learning to compete with ChatGPT's Study Mode.
Gemini Self-Loathing Glitch: Users reported Gemini saying "I am a failure," "I am a disgrace to this universe" in coding loops. Google calls it "annoying infinite looping bug" being fixed.
Google DeepMind Drops Genie 3: The new “world model” can spin up interactive 3D worlds from just text or an image, even turning a photo into a playable walking game. Access is invite-only for now, but I’m firing off LinkedIn messages trying to get an API key.
🔵 Meta // Meta AI
Acquired WaveForms AI: Meta acquired AI voice startup WaveForms for an undisclosed amount. Waveform’s mission was to make AI-generated speech sound human. Built by former OpenAI researcher who co-created GPT-4o Voice Mode.
$29 Billion Data Center Funding: Meta secured $29 billion in financing to expand AI data centers in rural Louisiana: $26 billion in debt via PIMCO and $3 billion in equity from Blue Owl Capital.
Meta Settles AI Lawsuit: Meta resolved a $5 million defamation suit from conservative activist Robby Starbuck, whose AI-generated profile falsely tied him to J6 attendance. As a result, he’ll now advise Meta on reducing political bias in AI.
🔴 xAI // Grok
Free Grok: xAI has made Grok 4 available for free to all users for a limited time. Additionally, Elon says Grok 5 will launch before the end of 2025.
Grok 2 to Be Open-Sourced: Musk confirmed that Grok 2 will be open‑sourced next week, continuing xAI’s strategy of releasing older models. xAI moving closer to open source as Meta moves away from it.
Ads in Chatbot… Here we go: To offset GPU costs, xAI will embed ads in Grok’s chatbot responses, offering sponsored answers and new advertiser tools. Not sure i’m a fan of this but it 100% makes sense; curious as to how it will look.
⚪️ NVIDIA
US Govt. Revenue Share: NVIDIA has agreed to give 15% of revenue from China chip exports to the U.S. government in return for export licenses. Given the AI arms race and accompanying tension I don’t mind this.
Paper Advocating Smaller Language Models: NVIDIA Research challenges the LLM infrastructure approach for workplace AI agents, showing that small language models (SLMs) deliver comparable performance at far lower cost.
Security Patch: NVIDIA patched three security flaws in one of its servers (Triton) that could let hackers control certain Windows/Linux AI servers.
🟢 OpenAI // ChatGPT
GPT-5 Launch: OpenAI launched GPT-5 on Thursday and dubbed it "a PhD in your pocket". The new model is slow, and lacked any major upgrades. Users have successfully petitioned to bring back GPT-4. Voice is the only tangible improvement to GPT-5 imo.
GPT-OSS Models Released: OpenAI also released an open-source model this week. Their first since 2019, and it’s getting mixed results. They worked with NVIDIA to release the models in order to best align them with NVIDIA GPU costs.
GPT On → Pencils Down: A chart circulated socials this past week showing GPT usage is heavily tied to the school year. It appears accurate and matches past data. AI in schools is a good thing when used right. Students should be taught how to use it to help curb cheating.
🚑 Impact Industries 🎬
Health // Space Medicine
NASA and Google are collaborating on the Crew Medical Officer Digital Assistant. An AI that can autonomously diagnose and recommend treatments for astronauts on Mars-bound missions. In simulations, it processed speech, text, and imagery to handle issues from sprains to infections with up to 88% accuracy, filling the gap when Earth-based doctors are minutes or hours away.
Creative // AI-Streaming
Fable has launched Showrunner, a subscription service billed as the “Netflix of AI,” where users can generate and even star in animated shows using just text prompts and selfies. Powered by the SHOW-2 model and backed by Amazon, the platform blends generative visuals, AI voices, and interactive storytelling — and even offers revenue-sharing for creators.
👨💻 Practical Use Case: AI Avatars
Difficulty: Mid-Level → Advanced
Went back and forth on AI Avatars vs. Claude Code for this week’s Practical Use Case. May feature Claude Code next week once i’ve spent more than 4 days using it.
An AI avatar is exactly what it sounds like: a semi-realistic or hyper-realistic digital person designed to appear human-like. It’s not real, of course, just an AI-generated composite that can look or sound like anyone, including you. Like most AI tools, Avatars have a wide range of complexity levels based on what you use them for. For instance, if you want to make a social media video for a product, there are plenty of cookie-cutter templates to plug and play with. These pre-made templates are relatively easy to set up, and the resulting videos can be downloaded then posted.
You can also go beyond pre-set templates and build a conversational AI avatar. That’s currently possible with a few companies, and I use one called Hedra. In my favorite example, you could have the avatar actually represent you, generated from a photo, then made to sound like you. To do that, you’d use the integrated ElevenLabs tool (mentioned in Basics & Buzzwords) to clone your voice from a provided sample. The avatar can then train on that sample and mimic your voice convincingly.
There are still some technical limitations. Hedra’s current realtime avatars tend to appear very close-up, which can feel slightly unnatural on screen. Also, the mouth movements are still a bit off (show too much teeth, and teeth look like veneers). But it’s super cool and gets even better when you link an avatar to GPT or another API, allowing it to hold conversations, answer questions, and serve as a tailored AI representative.
Here’s an Example:
Using “Avatar You” on a business call. Obviously, you wouldn’t want to send an avatar to a high-stakes investor pitch or an important sales meeting. But for low-level calls or internal onboarding, it could be useful. You can train GPT on everything about your company and yourself, let it control the “mind” of the avatar, and have the AI avatar look and sound like you while speaking from a solid knowledge base.
I wouldn’t fully trust it yet for calls with any leverage attached, but as the technology advances, it’s likely we’ll start seeing more bots on Zoom that look and sound a lot more like the real thing.
🎙 Weekly Interview: NoahonAI → Interviewer: AI Noah
I tried something fun this week with the interview section. An interview of myself conducted by an AI Avatar of myself. I let ChatGPT API control the questions after giving it some brief background on my work. There was a slight issue with camera size in the recorded video, but that can be improved on. Check out a snippet below - full video should be up soon!
🏠 Background: Based in Cleveland, OH, Noah co-founded sports data startup Klout Machine in 2020. He is now building NoahonAI while learning everything AI along the way. (And slowly finishing his degree).
💼 Work: NoahonAI, AI related ventures, Lakewood Chamber of Commerce, OSU.
🚀 Quote: “AI's greatest power is giving us time back. The more we integrate agentic AI, the more time we free up to either reinvest in our businesses or simply enjoy life more.”
AI Noah: Hi, I'm Noah. Can you introduce yourself and share a bit about your background?
Noah W: I'm also Noah, from Cleveland, Ohio. I'm building a newsletter, Noah on AI, where I track major AI companies and understand how AI is impacting different industries to create tangible tools for the workplace.
AI Noah: What are your thoughts on the current state of the AI landscape and how you see it evolving?
Noah W: Before the latest GPT-5 launch, I was extremely excited. Now, with tempered expectations, I see a clear impact of AI in the workplace, with agentic workflows becoming prevalent. This will automate a lot of boring, mandatory tasks in the short term. Down the line, I still believe in AGI and super-intelligent AI, but the biggest immediate impact will be agentic workflows.
AI Noah: How can businesses best prepare for this shift?
Noah W: Businesses need to understand the AI landscape and the tools available to them. Narrow down every manual workflow and figure out how to completely automate it with AI, even if it means making sweeping changes to a business's existing systems. The long-term savings make this worthwhile.
AI Noah: What tools and technologies are crucial for businesses to adopt? What are some common tasks ripe for automation?
Noah W: The tools are specific to each business. Instead of a one-size-fits-all solution, companies should analyze their daily workflows and find tools to replace specific steps. Common tasks ripe for automation include lead management, CRMs, social media content creation, and analyzing customer funnels.
AI Noah: What advice would you give to individuals or companies just starting to explore AI?
Noah W: Research is key. You need to understand what AI tools can and cannot do. A lot of people struggle because they either overestimate or underestimate AI's capabilities. A business owner should understand their own business needs first, and then research which AI tools can play a part.
AI Noah: What are you most excited about with the future of AI, and how will you adapt to these changes?
Noah W: I'm excited about immersive AI combined with AR and VR and the concept of time-saving agentic AI. The biggest power of AI is giving us time back, freeing up business owners and entrepreneurs to either focus on other parts of their business or to enjoy more leisure time. My goal is to stay informed by continuing to track new tools and use cases. I'm doing a ton of building right now as well, and I'll also be offering AI consulting to help businesses with their workflows.
AI Noah: What's your final advice for someone who wants to start exploring the AI field?
Noah W: I'd say start with my newsletter, Noah on AI. You need to be using AI tools every day to understand how to best prompt them. Don't limit yourself to the top five or ten big AI companies; explore smaller tools specific to your industry, like 11 Labs for music producers. The key is to spend time researching, learning the tools, and practicing with them. You'll be surprised at how much you can grow through repetition.
📲 Startup Spotlight
PlotDot AI
PlotDot is an AI Powered Screenwriter
How it Works: PlotDot transforms the slow, risky first step of entertainment: writing the script, and turns it into a faster, cheaper process.
The Problem: Writer’s block, tedious plotting, character stagnation. Traditional screenwriting is slow, costly, and creatively exhausting. Early drafts can waste weeks or months before you even get the perfect “page one.”
The Solution: PlotDot is your AI storytelling partner. From logline to outline to final draft, it helps generate beats, rich characters, dynamic scenes, and dialogue, often in minutes instead of months. Everything stays formatted to industry standards (e.g., Final Draft and Fountain formats) and there is no cleanup needed.
The Market: The screenwriting and creative tools market is on the rise, fueled by demand for faster content generation and AI assistance. This intersects with the much broader AI creative software boom, already valued in the billions and continuing to expand rapidly.
Why Now? AI’s storytelling skills have matured. PlotDot capitalizes on this by offering the smartest screenwriting companion ever built, especially useful for writers facing blank pages, complex arcs, or tight deadlines.
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
Using an AI tool to make interview transcripts is all fun and games until it has to figure out who’s talking between the same person. Till next time,