Join institutional investors, analysts and strategists from the world's largest banks: Subscribe today
Already have an account? log in
Google > OpenAI: A Grand Reset Of The AI Chessboard?
If you want to unlock the mysteries of this week's at times confusing US equities trade, look no fur
“A.I.” isn’t really artificial intelligence, it’s just advance Search. Google on steroids.
Want some code, Gemini can check a lot of references and reference them to give you some. There’s no critical thinking or desire going on behind the scenes. It’s just search that can apply findings.
These companies are reinventing the wheel and hoping the data center expenses are less than the human software costs. Adding “artificial” doesn’t change the intelligence part, which is what humans have been providing for thousands of years, to various degrees anyway. This AI isn’t even that…
This race between competing LLMs will probably be endless. One dominant model will rule the roost for six months until someone comes out with a marginally better model.
Some of you may remember a similar dynamic in the early days when stents in the early days. I doubt things are all that different when it comes to AI. Apple seems to be the only mega-tech vendor which recognizes this.
Agreed, it’s all PhD math stuff, it’s all about product execution at this point.
I’m flabbergasted by the market’s continual over-reaction to incremental improvements. Google carefully leaked the “model card” for Gemini 3, which showed its performance on a number of benchmarks to be superior to the rest. Shouldn’t observers know by now that benchmarks in computer world are nearly meaningless as an indicator of actual product quality? How are we still falling for the same numbers game and dog-and-pony show antics that let Bill Gates and Steve Jobs peddle hype for a quarter century?
I can conclude three things from the constant spectacle:
1. We want to believe in the next great thing.
2. The bubble has got everything inflated, even hype.
3. Many market actors don’t engage directly with the money-making, job-displacing business end of this tech.
Building a superior user experience with better results (more jobs made redundant) at a lower operating cost than the next guy will choose winners in this race, long term, same as any other innovation. Model cards and hyperbolic singularity tweets just don’t capture those viability details!
Agreed. The performance ratings are at best, a snapshot in time. AI requires very sophisticated data mining and processing. So access to proprietary data is a very substantial advantage. Google has decades of data from Search and from YouTube. And as physical AI evolves, Google WayMo, Nvidia, and Tesla have vast amount of data for development of autonomous driving, industrial automation, robotics etc. AWS has decades of proprietary data about consumer preferences. Facebook/META has more than a decade of proprietary social media data and so does ByteDance. What proprietary data does Open AI have beyond the past 3 yrs from Chat GPT queries and the public Web ? Open AI is in an extremely competitive arena of American and Chinese frontier models. Its totally dependent on private investor funding and debt financing and isn’t projected to break even till 2030 assuming a torrid rate of growth which is now at least questionable, rather than inevitable. Open AI CFO floating a trial balloon about federal guarantees to “backstop” their lenders/investors on a Bloomberg interview 3 weeks ago was the tell, and the proximate cause of the AI stock market euphoria suddenly becoming flimsy. While Google, META, AWS actually have existing highly profitable and sustainable enterprises to support their CapEx, All the hundreds of billions$$$ of CapEX deals announced in the past 2 months from the non-profitable and debt financed, are the stuff of dreams, as demonstrated by the plummeting share prices of Oracle and the Neoclouds.
Underscores how little durable advantage any model maker has. Incrementally bullish for the hardware names.
Gemini 3 is definitely a massive upgrade. I understand that its image processing capabilities alone put it on the top of the mountain. The new Antigravity IDE seems to be pretty good but I can’t say Gemini Pro 3 is better than Claude Sonnet 4.5. I used it exclusively for a few days and found myself back in Claude Code by the weekend. It’s better than Codex though. At this point I don’t see a reason to continue paying Open AI. Between Claude research/Code and Gemini 3 I have the capabilities I need to execute on pretty much anything reliably.
(If you haven’t tried Claude Research yet, I would suggest throwing it a detailed and difficult to source prompt. My last prompt came with 756 sources and took just over 15 minutes to run. Graduate level work)
Given that Open AI seems to have peaked and according to some folks, jumped the gun, I don’t know how they survive unless their backers sell a “too big to fail” narrative to the government.
I have been exploring Gemini 3 pro in the AI Studio interface the day after it came out. For me, it has been massively superior to GPT5.1. So much so, I subscribed to Gemini Pro within an hour and have been using it extensively for days. Watching its thinking log has been as illuminating as the outputs. And I am only doing analysis on markets. I haven’t even tried the image stuff. And I haven’t “vibe coded” anything yet. In my experience the GPT coding has been hit or miss in my applications. I did experience a few errors, but I am very impressed so far with Gemini. It’s amazing how early we still are in AI capabilities.