Was there a new foundation model released today?

Has clearing the browser cache/cookies ever solved a problem other than making the next week browsing suck?

I’ve only gone truly negative once on a company’s solution once that I remember. It was VMware Cloud. But the reason I went negative because I was still a practitioner and the product let me down when my son participated in a hackathon at VMWorld. It makes me wonder what makes @SemiAnalysis_ so angry at $IREN. Analyst firms rarely go scorched earth on a product. Truly negative takes like this don’t make it through a mature organization filter. I was surprised at how big of a platform I had developed in the space. SA must have really been offended by $IREN managed cluster product.

My brother @CoachTtownsend owns a @maaco North of Milwaukee in Menomonee Falls. He's owned it for four years and came to the realization, it's a Maaco. What does that mean? It means that they are really good at making a daily driver a different color. Not at restoring an antique or show car. This is an important thing to realize, no matter your business. When you are good at something, embrace it and don't apologize for it. When someone asks you to do something you aren't good at, be upfront. There's a market for the thing you do.

"We removed the compilers from the servers to make them more secure. Now we're racing to put the compilers back with tools that can generate their own code and use them." - Kelsey Hightower via LinkedIn.

One, look at the price. Two, look at the shape. Three, look at how that software has drastically improved. If @Apple decided to offer a server class system with dual power supplies and 1 or 2U configurations, they will make noise in enteprise inference.

I have a call with a potential customer. Should my coffee be ice or hot?

OpenAI’s most consequential contribution may not be GPT. It may be creating the API abstraction that separated AI applications from AI infrastructure. The OpenAI API has become a de facto inference ABI. Vendors can fight over CUDA, ROCm, TPUs, memory architecture, inference engines, and accelerators underneath it while software developers increasingly just change a model name and endpoint.

There’s a reason that @NVIDIA RTX 5090 and RTX 6000 are sold out. I’m almost certain you can walk up and purchase a GB300 workstation. It’s not the right shape for the data center. I don’t mean the case. I mean the memory configuration. These desktop/workstation class cards are the workhorses of enterprise AI. Fast enough inference for models that fit on up to 8 of these cards in a server is the sweet spot. 8 may be on the high side. That’s why you’ll see @Apple sell as many of the M5 Ultras with 96/256/512GB of RAM that they can make. Benchmarks show the 5090 just edging out an M5 Ultra. The unified memory is both a strength and weakness. It doesn’t matter. I expect to see Mac Studios in data centers everywhere.

Clustermax 3.0 This has become an important instrument. My only fear is that the narrow audience for this instrument is applied too universally. Companies with 7 to 10 million dollar compute budgets isn’t representative of the greater market. Managed clusters are an important capability but isn’t going to be the majority of the market. newsletter.semianalysis.com?…

If The Machine had succeeded before modern deep learning scaled, would we have discovered a different model architecture that looked just as inevitable to us today? An unanswerable question. HPE would like that one to run back. Memory-driven compute was the right concept.

Wow... 256GB M5 Ultra deliveries are showing next year as the next available date. I think I'll hold on to my October order at least until I see what the ship date would be for the 512GB system.

I applied for a job as <insert here>

I've become the 50+-year-old dude with a six-year-old pickup truck. There's not a new car that excites me.

Apples and almonds. That is all.

Some of our best recent hires have been found here on X. So I’m going back to the well here. We are hiring Tech Analysts at Futurum. Looking for AI-pilled, tech-obsessed, data-loving, social media crazed analysts across Semiconductors, Infrastructure, Cyber, Developer/DevOps, and broadly across the AI sphere. Analysts that join us will be part of a team that works closely with the largest technology vendors, financial services firms, and enterprise users in the world. The work will include (but not limited to) -Product Research -Market Modeling -Testing and Validation -Evaluations and Assessments -Attending Vendor Events -Public speaking (TV, Podcasts, Radio) -Sharing your views on X -Lots of other fun stuff We are fast growing firm that is building technology, modeling important markets, providing market education, and we are always innovating across our businesses that include @TheFuturumGroup, @FuturumEquities, @ETRnews, @TechFieldDay, @TechstrongTV, @VisibleImpactUS, @TheSixFiveMedia, and @Signal_65 Experience from various backgrounds will be considered. Entrepreneurial, highly motivated, intellectually curious, and overly opinionated (with work to back it up) will receive highest consideration. Reply below or email website@futurumgroup.com and feel free to pass along or share with your AI-pilled super smart friends that may be perfect for this gig.

My M5 Ultra with 256GB isn't scheduled to come in until Oct. 20th. I haven't decided whether to cancel the order yet. The YouTube reviews should be hitting any minute.