Noah on AI: Issue 21 (RECAP)

Welcome to the 21st edition of NoahonAI.

What’s included: company moves, a weekly winner, AI industry impacts, practical use cases, and more.

Hope everyone enjoyed their Thanksgiving!

Just a recap and some thoughts on the NVIDIA5 for today’s newsletter.

Definitely worth looking back at some of the past Practical Use Cases.

Links are scattered in the text but each one will take you to the original issue the story was published in.

I changed the link color just for today’s issue to blend in a bit more because there’s a ton of them.

Interview #5 might be one to check out for those who haven’t seen it yet.

Finished up the 1st sponsorship deal with VERN AI - make sure to check them out if you didn’t get the chance!

Next week we’ll cover all updates since last Tuesday.

NVIDIA isn’t going anywhere: NVIDIA has sat atop the AI race leaderboard since I started tracking back in July. They are positioned as the lead supplier in a race with near unlimited demand. While they have dealt with their fair share of issues on sales restrictions, China specifically, it has not slowed them down one bit. NVIDIA appears to be extending into a new international market while developing or updating their tech every week. Their ability to react and innovate with change makes them near indefensible, even as TPU’s and competitors move further into the picture.

I’m becoming a bit concerned: ChatGPT was the top product in AI, and remains so on the consumer front. It has the first mover and has continued its growth over the past 4.5 months. However, I’m noticing a trend in my GPT usage where speed to update is overtaking quality. Minor bugs, context failure, and other menial issues hurt the product in my eyes. GPT-5 was also a bit of a let down. I’m still bullish on GPT, they have the best features, including some that most others don’t even have. BUT, it’s important they focus on making the baseline product and features work to the best of their ability before trying to iterate more. Features are great, but the hype fades fast if they’re not usable.

Seems like they’re doing everything right: Google was slow to the AI race until Cofounder Sergey Brin returned back in late 2023. Since then, they’ve been on fire. The image and video models have been excellent. The API tools, specifically those around RAG seem very impressive, and the Gemini 3 launch was widely regarded. The consumer facing LLM is still a bit unimpressive, but with their Google Deepmind arm, API features, Multimodal tech, and perhaps a venture into robotics, it’s only up from here. Not to mention training all of these models on their own custom hardware. Very impressive 4.5 months.

My new #1 LLM: Claude is awesome. AI assisted coding has become a huge market, and Anthropic has always had the best models for that use case. However, they didn’t stop there. Anthropic has firmly staked its claim in the enterprise arena and remains the top business model. From a usage standpoint it has the best mix of conversational prowess and intelligence. If I had to ask one LLM a question where I needed a straight answer it would be Claude Opus 4.5. I’m not sure if coding, enterprise, or both will be the dominating use case when its all said and done, but outside of Google, Anthropic has been the most impressive model since I started tracking the race.

Turning the corner: I’ve had the same message for xAI since I started writing about them so there’s no need to change it now. The model itself is solid, but it has been in the wrong place directionally. Compared to the above LLM’s, it’s only advantage use case is social media. It has impressive benchmarking numbers but, from my time using it, and for those I’ve spoken to, the numbers don’t represent the full picture. xAI has a ton of potential in gaming, physics, and with its connection to Tesla, Robotics. I think we’re gonna see more progress on those fronts in the coming months, especially as Tesla moves forward with its humanoids.

What in the world is going on at Meta: I was excited about Meta’s growth over the past few months, especially after they went on a massive hiring spree to bring in a lot of top AI minds. However, since then, it does not seem like anything has come of it. Open source models are their current best use case, but that could be killed tomorrow if a better company open sources one of theirs. The biggest product drop since the tracker started was AI glasses that no one uses, or Vibes, the AI slop video app. The spending is going to be there, and the people are in place, so i’m hoping Meta finds a way to build better stuff as we head into the new year.