B (4:10)

Man, the hits just keep coming for OpenAI this week, don't they? And hey, for Nvidia 2. Amazon yesterday launched Trainium 3, saying this new AI chip is four times faster than Trainium 2 and can cut AI training and operating costs by up to 50% compared to equivalent GPUs, quoted the Journal. The main advantage at the end of the day is price performance, said Ron Diamant, an AWS vice president and the chief architect of the Trainium chips. He added that his main goal is giving customers more options for different computing workloads. I don't see us trying to replace Nvidia, diamant said. Yeah, but perhaps more pertinently, Amazon also announced Trainium 3 Ultra Server, a system powered by those 3 nanometer Trainium 3 AI training chips, and teased Trainium 4, which it says will work with Nvidia's chips. Quoting TechCrunch, AWS says the system is more than 4x faster with 4x more memory, not just for training but for delivering AI apps at peak demand. Additionally, thousands of Ultra Servers can be linked together to provide an app with up to 1 million Trainium 3 chips 10x the previous generation. Each Ultra server can host 144 chips, according to the company. Perhaps more importantly, AWS says the chips and systems are also 40% more energy efficient than the previous generation. While the world races to build bigger data centers powered by astronomical gigawatts of electricity, data center giant AWS is trying to make systems that drink less, not more. It is obviously in AWS's direct interest to do so, but in its classic Amazon cost conscious way, it promises that these systems save its AI cloud customers money too. AWS customers like Anthropic, of which Amazon is also an investor, Japan's LLM, Karakuri, Splash Music and Descartes have already been using the third gen chip and system and and significantly cut their inference costs, Amazon said. AWS also presented a bit of a roadmap for the next chip, Trainium 4, which is already in development. AWS promised the chip will provide another big step up in performance and support Nvidia's NVLink Fusion high speed chip interconnect technology. This means the AWS Trainium 4 powered systems will be able to interoperate and extend their performance with Nvidia's GPUs while still using Amazon's homegrown lower cost server rack technology. It's worth noting too that Nvidia's CUDA has become the de facto standard that all the major AI apps are built to support. The Trainium 4 powered systems may make it easier to woo big AI apps built with Nvidia's GPUs in mind to Amazon's cloud. Amazon also released its second gen Nova AI models including Nova Lite, Nova Pro, Novasonic and fully multimodal reasoning model Nova Omni to limited customers. Quoting Wired, Amazon detailed two improved large language models, Nova Lite and Nova Pro, a new real time voice model called novasonic and a more experimental model called Nova Omni that performed a simulated kind of reasoning using images, audio and video as well as text. The new models are being made available today to a limited number of customers. More significantly, given the importance of its cloud business, Amazon is releasing a tool called novaforge that will let customers create specialized frontier models by adding their own training data to unfinished versions of the Nova 2 Lite and Pro models. It is already possible to fine tune off the shelf AI models like Google's Gemini and OpenAI's GPT, but Amazon's approach lets customers add data at various stages of model training, including the process of building the base model, a stage known as custom pre training that is normally reserved for large AI labs. Everyone is looking for a frontier model that's an expert in their domain, rohit Prasad, who leads Amazon's AI efforts, told Wired ahead of today's announcements. Prasad says that Amazon developed the technologies behind novaforge to empower internal teams, including those developing Alexa and AI agents, to build custom models. This is essentially a new open training paradigm, he says. One customer that has tested the approach is Reddit, which used Nova Forge to create a custom model to identify content that breaks the platform's rules. Fine tuning a conventional model would not work, says Reddit Chief Technology officer Chris Slow, because most models are designed to avoid offensive or violent content entirely, meaning they would refuse to analyze some materials. SLO says that custom pre training combined with conventional fine tuning produced a frontier model that is expert at understanding and using Reddit. End quote. Finally, Amazon also has three Frontier AI agents the Kiro Autonomous Agent, the AWS Security agent, and AWS DevOps agent, each focused on a different aspect of software development.

B (8:53)

One of the longtime selling points of self driving cars has always been look in aggregate, the robots are probably much better drivers than humans are, and in an analysis of Waymo Data covering nearly 100 million driverless miles in four US cities seems to bear that out. Waymo cars have far lower crash rates per million miles than human drivers do. This is quoting from an opinion essay in the New York Times from neurosurgeon Jonathan Slotkin. When compared to human drivers on the same roads, Waymo's self driving cars were involved in 91% fewer serious injury or worse crashes and 80% fewer crashes causing any injury. It showed a 96% lower rate of injury causing crashes at intersections, which are some of the deadliest I encounter in the trauma bay so far. Other autonomous vehicle companies don't report or report incomplete data. Waymo, by contrast, published everything I needed to analyze the data crash statistics with miles driven that allow accurate comparison to human drivers in the same locations. If Waymo's results are indicative of the broader future of autonomous vehicles, we may be on the path to eliminating traffic deaths as a leading cause of mortality in the United States. While many see this as a tech story, I view it as a public health breakthrough. In medical research, there's a practice of ending a study early when the results are too striking to ignore. We stop when there is unexpected harm. We also stop for overwhelming benefit when a treatment is working so well that it would be unethical to continue giving anyone a placebo. When an intervention works this clearly, you change what you do. There's a public health imperative to quickly expand the adoption of autonomous vehicles. More than 39,000Americans died in motor vehicle crashes last year, more than homicide, plane crashes and natural disasters combined. Crashes are the number two cause of death for children and young adults, but death is only part of the story. These crashes are also the leading cause of spinal cord injury. We surgeons see the aftermath of the 10,000 crash victims that come to emergency rooms every day. The combined economic and quality of life toll exceeds $1 trillion annually, more than the entire US military or Medicare budget, end quote.

B (11:15)

The last thing you want as an IT professional is for an auditor to catch something before you do. That's why Yeshid automates provisioning, deprovisioning, access reviews and SaaS discovery through the lens of security and compliance, giving IT leaders full visibility, consistent policy enforcement and an audit ready posture. Yes ID delivers advanced IAM automation without forcing teams into legacy identity providers. Whether you use Google Workspace, Microsoft 365 or Okta, Yeshid integrates directly. Plus, Yeshaid's flexible group and role model gives it precise control over access provision automatically when users join a group or only on request. That reduces license waste and maintains least privilege. Learn more@yeshhid.com Techbrew that's yeshid.com Techbrew.

B (12:12)

To say compliance is complicated is an understatement. Constantly worrying about SOC2, ISO, HIPAA, CMMC, FedRamp and more can leave your head spinning. Even worse, a misstep can get pretty costly for your startup. That's why Delve is designed as an AI native compliance platform. Delve uses AI agents to handle headaches like taking repetitive screenshots for you, monitoring your tech stack for security gaps in real time, auto filling security questionnaires in your browser or in CSV, and creating secure data rooms to send to prospects and auditors. Their team can personally work with you in Slack to get you 100% compliant and manage your whole audit for you. Over 1000 of the fastest growing companies get compliance done in Delve, including lovable, bland, micro one instantly and 11x for listeners. They're offering an exclusive $1,000 discount on any compliance framework. Check out Delve Co Morning Brew and start using AI for compliance. That's Delve Co Morningbrew.

B (13:20)

In the great debate about the possibility of an AI bubble, one of the angles to consider is are any of these tools compelling enough for people to pay up for. Well, Microsoft especially continues to give us some concerning data points, quoting the information Executives at Microsoft and other enterprise software firms heralded 2025 as the year artificial intelligence would be capable of automating tasks that involve multiple steps, such as generating dashboards based on company sales data process. But as the year comes to a close, Microsoft has lowered expectations for how quickly it can get customers to spend money on these newer products, known as agents. Multiple Microsoft divisions, for instance, have lowered how much salespeople are supposed to grow their sales of certain AI products after many of them missed sales growth goals in the fiscal year that ended in June, according to two salespeople in Microsoft's Azure cloud unit. It's rare for Microsoft to lower such quotas for specific products, the people said. The change shows how Microsoft is adjusting to resistance from companies to pay more for AI. Corporate customers have complained over the past year that it's hard to measure the savings from using the technology for tasks like drafting reports on customer spending and sales leads, and that it can be difficult to get the AI to work perfectly in cases where small mistakes can be costly, such as automating finance and cybersecurity tasks. That said, AI has been a major boon to Microsoft's business. That's largely thanks to new spending by AI firms such as OpenAI, which has projected it would run about $15 billion worth of cloud service from Microsoft this year, as well as Microsoft's sales AI software such as its 365 Copilot workplace tools and GitHub Copilot coding agent. Microsoft itself and other large tech companies also have gotten productivity boosts from using AI tools internally. But getting traditional businesses to boost their spending on advanced forms of AI hasn't been as easy. For instance, private equity fund Carlyle last year started using Copilot Studio, a Microsoft product that lets companies develop AI to automate tasks like summarizing meetings or drafting financial models based on Excel spreadsheets without needing to write any code. But in the months after Carlyle started using the tools, its representatives told Microsoft they were having trouble getting the AI to reliably tap data from other applications, such as Salesforce's Customer Relationship Management app, which was necessary for some of Carlyle's automations, according to someone with direct knowledge of the situation and a second person briefed on it this fall, Carlyle reduced the amount it was spending on the tools, the people said. Microsoft isn't the only firm adjusting expectations for revenue from AI agents that automate complex tasks. OpenAI for instance, recently recently lowered its projections for AI agent revenue by $26 billion over the next five years compared to earlier numbers. OpenAI expects to make up for that by growing subscription revenue from ChatGPT, which includes some agents such as Deep Research for creating research reports, as well as from new products, which could include selling ads on ChatGPT, according to the projections, though CEO Sam Altman said Monday he was putting off work on ads as well as agents related to health and shopping as the company focuses on fixing problems involving ChatGPT. The challenge of increasing revenue from AI agents been particularly acute at enterprise software firms such as Salesforce that, unlike Microsoft, don't have the benefit of owning a large cloud server business. Salesforce and other firms such as ServiceNow have been offering steep discounts to customers that try new agent products to automate workplace tasks like closing IT tickets or onboarding new employees. Other firms, including Amazon Web Services and Anthropic, have been pouring resources into helping customers set up AI applications to run properly, similar to the way consulting firms help their clients. It's not the first time a large cloud and software provider wider adjusted expectations regarding how much enterprises will spend on AI, even as sales of chatbots and coding models have surged. Last year, for instance, Google and Amazon tamped down expectations for enterprise AI sales after companies didn't start paying for new AI tools as quickly as they had hoped. But there may be sunshine at the end of the AI agent tunnel. Brian Spanzwick, chief information officer at cybersecurity firm Cohesity, said his company has been testing AI agent features from Copilot Studio and Foundry that can write reports about its customers for salespeople based on internal data Cohesity has on them as well as publicly available information. While agents have failed to deliver on their promise so far, Cohesity is writing code to better connect them to data sources from other applications with hopes of making them work over the next few months. That's what I'm telling my board, he said. Talk to me in three months and I'll show you the roi, end quote.

B (17:59)

Finally today, HBO Max recently released a 4K upscaled re release of the TV show Mad Men. But, well, not great. Bob quoting the Hollywood Reporter. For starters, the Mad men episodes in 4K on HBO Max were listed out of order and mislabeled, which likely made series creator Matthew Weiner an absolute madman. But that's nothing compared to an actual edit Oopsie in the seventh episode of season one in Red in the Face what the episode is supposed to be titled, but was not on 4K. HBO Max a drunk they were always drunk. Roger Sterling, played by John Slattery, vomits up oysters in the office in front of prospective clients. In the initial 4K stream that released on HBO Max, a Mad Men crew member was very visibly operating a barf hose behind Sterling in full view of the camera. There was even a second staffer in frame supporting the pukehose tech. That guy would have been a great friend in college. So what happened here? Lionsgate delivered the wrong 4K file to HBO Max. The Hollywood Reporter was told the non 4k files were fine as delivered. At the time the story was first published, Lionsgate was working on getting HBO Max the correct files. THR was told they were then in the process of being swapped out at around 10am PT on Tuesday, December 2nd. 2nd. Lionsgate did not immediately comment on the fail. HBO Max referred THR back to Lionsgate. HBO Max has been heavily promoting its acquisition of Mad Men reruns and especially touted how Mad Men on HBO Max would be the series first time ever in 4K. The launch was probably not quite how executives pictured it. Literally. End quote.

B (19:49)

I forget if I mentioned this already, but I've been busy churning out even more rad history episodes. I did one on New Jack Swing. I just researched a three part episode on Hair Metal yesterday and if I get time to edit this afternoon, I'll have an episode out on the US invasion of Granada in 1983. This is what always excited me about this project. The ability to have a range of topics as broad as Guns N Roses.