Podcasts about Ampere

This week, Paul and Sean begin with the recent development that several prominent SVOD platforms are exploring adding a free tier to expand their subscription funnel and sell increased advertising revenue. Next, they discuss a study released by Ampere that stated US and UK individuals spend more time consuming media than doing anything else (including work or sleep), and break down the categories and findings. Finally, they dissect a recent report from the WGA West about writers' earnings over the past 10 years and wonder whether peak TV will return. Learn more about your ad choices. Visit megaphone.fm/adchoices

Watch the full episode on YouTube:We first covered Baseten last year when DeepSeek mania was at peak hype. Now they have raised a monster $13B round and become one of the new cohort of AI Infra decacorns that are (with Nvidia, Intel, and the semis complex) chief beneficiaries of the Inference Inflection. We return to Baseten at the peak of the 2026 edition of Open Weights debate. Ali has published a viral breakdown of Kimi K3:And since you last saw him, Philip has spoken at AI Engineer and written the definitive book on Inference Engineering spotted all over SF:Three years ago, inference engineering barely existed as a category.Today, it is one of the most critical disciplines in AI. Inference engineering inherently tackles a different question than standard model training: “How do you turn those weights from training into a product that is fast, reliable, and affordable at scale?” Focusing on these creates an entirely new optimization problem.In one recent GLM-5.2 experiment, quantizing more of the model actually preserved its benchmark quality while increasing throughput by 20%, because the errors introduced in different layers could cancel each other out.Inference is no longer just the final step after training. It is becoming its own engineering discipline, with its own research problems, infrastructure, and increasingly specialized roles.In this episode, Baseten's Philip Kiely and Ali Taha join swyx and Vibhu to explain what actually happens after a new open model is released and what it takes to turn “we generated a token” into a fast, reliable, production-ready API.We go deep on cache-aware routing, disaggregated prefill and decode, quantization, speculative decoding, KV-cache movement, model parallelism, GPU kernels, and the race to make frontier models up to 10× faster. Philip and Ali explain why inference optimizations can still produce gains of 20%, 100%, or even 200%; how quantization errors can cancel one another out; why identical weights can behave differently across clusters; and how Baseten grafted a Kimi vision encoder onto GLM-5.2 without changing the underlying language model.The conversation then expands beyond LLMs into NVIDIA Dynamo, mega kernels, Rubin, AI-specific chips, local inference, video generation, diffusion versus autoregressive models, and the enormous compute barrier to generating coherent long-form video. Finally, we explore the convergence of training and inference, continual learning through persistent KV cache, and the emerging loop where models help optimize the infrastructure that runs them.We discuss:* What happens when a 200,000-token request enters an inference system* Cache-aware routing and reusing previously computed KV cache* Why prefill and decode are increasingly handled by different GPUs* When dedicated deployments become cheaper and more reliable than shared APIs* How speculative decoding uses a smaller model to accelerate a larger one* Tool calling, structured outputs, and what LLMs actually do* What it takes to support a new open model on day zero* Grafting Kimi's vision encoder onto GLM-5.2* Retrofitting inefficient model layers with components from other architectures* Why models sometimes collapse into repeating the same token* How hardware, kernels, and race conditions create nondeterministic failures* Preserving model fidelity while making inference faster* How quantization errors can cancel each other out* Why inference optimizations still deliver gains of 20%, 100%, and 200%* How optimized serving can make a model up to 10× faster* NVIDIA Dynamo, KV-aware routing, and distributed model serving* Speculative decoding the speculative decoder* Why local AI is about making models less dumb while data-center AI is about making them less slow* Tensor, expert, and pipeline parallelism across GPUs* Hardware-aware model design, auto-tuning, and the case against mega kernels* Rubin and why inference is becoming a systems problem* Whether modern GPUs are evolving into programmable AI ASICs* Why enormous models like Kimi K3 require GB300-class hardware* Why open-source video generation still trails Veo, Kling, and other closed models* The quadratic attention bottleneck behind long-form AI video* Autoregressive video, real-time generation, and compounding quality drift* Why future video systems may combine autoregressive and diffusion architectures* Training for inference and inference for training* Continuous post-training, deployment, evaluation, and improvement loops* How GLM-5.2 helped optimize the kernels serving GLM-5.2 itself* Why faster networking could unlock dramatically faster decoding* Continual learning, KV-cache compaction, and persistent model memoryShow Notes* How to build a day-0 API for Kimi K3* 22580: From GPT2 to Kimi3, ExplainedPhilip Kiely* LinkedIn: https://www.linkedin.com/in/philipkiely* X: https://x.com/philipkiely* Inference Engineering: https://www.baseten.co/inference-engineering/Ali Taha* LinkedIn: https://www.linkedin.com/in/aliestaha/* X: https://x.com/waterloointernTimestamps00:00:00 Introduction and the 200K-Token Prompt00:03:18 Dedicated Deployments, Speculative Decoding, and Tool Calling00:11:26 Launching Production-Ready Open Models00:19:06 Model Retrofits, Failure Modes, and Nondeterminism00:28:22 Quantization and Canceling Errors00:32:15 The Race to 10× Faster Inference00:40:48 Dynamo, Speculation, and Local vs. Data-Center AI00:50:18 Model Parallelism, Auto-Tuning, and Mega Kernels01:00:55 Rubin, GPUs vs. ASICs, and Custom AI Chips01:10:03 Giant Models and the Limits of GPU Memory01:12:42 AI Video, Quadratic Attention, and Autoregressive Generation01:21:47 Audio, Images, and Diffusion Models01:27:32 Training, Self-Optimizing Models, and Continual Learning01:40:06 Closing ThoughtsTranscriptIntroduction: Baseten, Waterloo Intern, and Inference EngineeringSwyx [00:00:00]: Okay, we're here in the studio with Philip, old friend from Inference Engineering, the book, as well as Baseten and everything that you've done, you and I have done before, as well as Ali. Welcome.Ali [00:00:15]: Pleasure to meet you.Swyx [00:00:15]: Waterloo intern.Ali [00:00:16]: Waterloo intern, always.Swyx [00:00:17]: When did you get “Waterloo intern” as a handle?Ali [00:00:19]: As a handle? Oh.Ali [00:00:20]: I think the rebranding happened mid-March. When I saw it was open, I was like, “I have to take it. Up for grabs.”Philip [00:00:26]: The problem is that Ali is really good at his job and is not gonna be an intern much longer.Philip [00:00:30]: So we have to figure out who's gonna get the handle.Ali [00:00:33]: Well, I'll pass the torch over to the next intern.Swyx [00:00:34]: Oh, okay. It can be, like, you just pass it to another Waterloo grad.Ali [00:00:37]: To another Waterloo intern. No, bruh.Philip [00:00:39]: Yeah.Ali [00:00:39]: Intern.Swyx [00:00:40]: Intern, yeah.Ali [00:00:40]: And no.Philip [00:00:41]: You gotta get an intern from Waterloo.Ali [00:00:42]: Yeah, I've gotta get an intern from Waterloo.Swyx [00:00:44]: Right.Ali [00:00:44]: But they have to follow the path.Swyx [00:00:45]: Oh, it could, but it could come from Baseten, so it's like whoever Baseten gets from Waterloo.Ali [00:00:48]: Right.Swyx [00:00:49]: Has the title of Waterloo.Ali [00:00:50]: It stays in the ecosystem.Philip [00:00:51]: Exactly.Ali [00:00:52]: Halfway through the internship, you either get it or you're out.Philip [00:00:55]: You should also do, like, a big graduation ceremony where you change the handle.Ali [00:00:59]: Just say it.Philip [00:00:59]: For everybody.Swyx [00:01:00]: You guys are good at ceremonies, clearly. We had a nice launch of the book, very successful. But before we get into all that, I wanna start off with a fun question for you. Okay, you're an expert inference engineer. What happens when I send a long query, say two hundred thousand tokens into Baseten's inference? What's the process of query through GPU model routing, balancing, all that? What is all the stuff that we don't think about?Long Context Requests, KV Cache, and Cache-Aware RoutingPhilip [00:01:26]: With a long query specifically, the first thing that I'm gonna ask is, “Have you sent me this query before, or at least part of it?” and I really hope you have, because it's gonna be a lot easier for me and a lot cheaper for you. So the first thing that we're gonna look at is some cache-aware routing, where we're going to see, we probably have a number of instances, a number of replicas up serving whatever model you're hitting. We want to send this one to something with, number one, available prefill workers, and number two, ideally some cached input already there so that we can skip prefill on at least part of these two hundred thousand tokens. If you're doing two hundred thousand tokens, it's probably coding or a multi-turn agent or something where you would expect to have that cached. If you don't, we're gonna have to send it to a prefill worker. We've at least on certain models disaggregated prefill and decode, so you're going to have one set of GPUs that's solely going to process the input, create the KV cache, and get you your first token, and then that's going to be passed over to a separate set of GPUs, which is going to run decode. We're going to iteratively make those tokens. We're probably going to have some speculator model in front of that. I'm going to assume that you're doing coding, and because of that, our speculator model, which assumes you're doing coding, is gonna have a high draft token acceptance rate. If I'm wrong and you're asking me to summarize every Harry Potter book, it's gonna be slower. And then we stream that output to you and account for it, charge you, a couple of pennies and say, “Hey, would you like to send another one?”Swyx [00:03:04]: Except Baseten doesn't charge by pennies.Philip [00:03:07]: Well, yeah, we charge. I'm assuming that we're talking about the public model APIs. If you are setting up a dedicated deployment, then yeah, it's not pennies.Public APIs vs. Dedicated DeploymentsSwyx [00:03:18]: Yeah, one of the key differentiators when I was talking with Baseten initially was that people who want very high volume just need to rent by the box, ‘cause then it's up to you to figure out how to saturate the box.Ali [00:03:31]: And more often than not, it's, like, way cheaper if you're pushing, like, millions of tokens per hour, if you just pay per hour instead of pay per token.Philip [00:03:37]: Yeah, they do. I think that we've increasingly seen a lot of demand for the pay per token APIs, just because everyone wants to try open models, and then once they find a use case that's really sticky, then they move over to dedicated.Swyx [00:03:51]: Is there a best practice on when it's time to swap over?Philip [00:03:54]: Couple reasons. Yeah, reliability, that's a big one, right?Ali [00:03:57]: Like, if they have a very specific use case, they want you to train something specifically for them, like they want their own spec dec, for instance, for their own traffic.Swyx [00:04:04]: Spec dec is speculative decoding.Speculative Decoding and Custom SpeculatorsAli [00:04:05]: Speculative decoding, yeah.Swyx [00:04:07]: You have to explain.Ali [00:04:07]: Sorry. Like, speculative decoding is like, if you have a huge model, right? And so the model is going to be generating one token at a time every single turn, every single forward pass. So we attach, like, this little, like, parasite, like this layer that goes on top of the model, and this model just has to predict. It does three very fast autoregressive forward passes, and it will predict, like, three certain tokens, and then you do one forward stage over the entire original model in order to see if those predictions were correct or not, and then you accept them or you reject them. Now, this draft model is traffic specific, so if you, like, Philip said, if you're summarizing Harry Potter books, I can train exclusively that draft model on Harry Potter books, and I can guarantee you that I'm gonna accept the three tokens every single time. And so with that case, I increase your decode speed. I wouldn't be able to provide this to you if you're a shared endpointSwyx [00:04:53]: YeahAli [00:04:53]: ‘cause I have no idea if you're doing Harry Potter, if you're doing coding, if you're doing English. We don't know. Also, there was a thing in the book that mentioned that if they really cared about a specific threshold, chapter four, I think. Do you remember that?Philip [00:05:06]: Yeah. The things that you can do is you can set a specific, like, batch sizing, a specific, like, parallelism strategy if you're trying to optimize for, like, throughput versus latency. You can. Maybe a NVFP4 quant doesn't pass your benchmarks and you wanna run a model at higher precision, you could do that. There's just a bunch of reasons why you might wanna have your own endpoint and the biggest one, of course, just being, like, you don't have to deal with someone else doing a hundred million tokens of benchmarking traffic at the endpoint when you happen to be trying to serve your users.Swyx [00:05:40]: Yeah. I think one thing that is. That is a classic journey. Like, it's people is asking the, what happens when you type Google into the browser. Tool calling, is that just, you're generating JSON or is there more complication beyond that?Tool Calling, JSON, and Structured OutputsAli [00:05:58]: Certain customers that we have, they have their own post-trained models, and so they demand a tool calling that's not just, like parse a file or go find the weather. It's something that's very specific and you have to do post-training on this. And if the post-training on the model is not good or if the quantization after the post-training to get the inference to be fast, the model will struggle reading the JSON file and reading the tool calling. But it doesn't require its own like sandbox. It's not like it's going to use that tool calling to like escape a sandbox or like it doesn't have to be contained. It can just be a normal dedicated deployment. The challenge with tool calling more and more seems to be that the companies want certain tool calling which is a very sensitive thing to train. And because you're dealing with all of the JSON outputs, if it doesn't like close the end of the request in a very certain manner, you end up with a model that did the tool calling and like the thinking and so as a result of that, it didn't see the result and just hallucinated the result as it decoded. That seems to be the most challenging thing with tool calling, not really the sandboxes model.Philip [00:06:56]: Yeah, that's a challenge on the training side and then on the inference side, there's work that you can do to scope the possible output. So we published this at this point close to two years ago, the solution to this problem which is you make a state machine and you use that to constrain the output to a specific format. So this is the structured output problem. If you remember backSwyx [00:07:27]: Yeah, the specific grammar is,Philip [00:07:29]: Yeah, exactlySwyx [00:07:30]: GML had this thing.Philip [00:07:31]: Yeah. So it's like the old-school “make sure this is only JSON”, return only JSON orSwyx [00:07:38]: YeahPhilip [00:07:38]: Grandma's gonna die type of prompts.Swyx [00:07:39]: Is it BNF grammar? At some point OpenAI had released a thing that was like, yeah, if you want to constrain your output, write BNF grammar, back as NOR.Philip [00:07:47]: In our inference system, it's just a specified output format. And you get the guarantee that your output's gonna be structured along that format. And so applying that to tool calls can like help cut down on. You can still call the wrong tool or call no tool. It doesn't solve the certainty problem but it at least solves the output structuring problemSwyx [00:08:10]: YeahPhilip [00:08:10]: Within tool calls.Swyx [00:08:12]: And MCP is just another form of tool, right.Philip [00:08:14]: Yeah, exactly.Swyx [00:08:15]: As far as there's no special thing there.Philip [00:08:16]: The thing I'm always like explaining to people is the LLM is not capable of doing anything. It's only capable of making suggestions of what to do and then if those suggestions are formatted in a certain way and applied to a system that knows what to do with them, then an action occurs.Swyx [00:08:32]: Yeah. Part of the fun stuff is, this is solved outside of tool calling too. Like in an agent loop if the output is not correct or you're right, like reasoning, tool calling was done in the reasoning trace, just be like, “Oh, I don't know what to do. Let me just try again.” And it might get there after a few tries. And on your point of training, sometimes this is harder in smaller models, so you don't have the same exact quality outputAli [00:08:56]: Right.Swyx [00:08:57]: When you just swap from a big model, right?Ali [00:08:59]: Yeah. I will say that, before, I think we need to go back to inference engineering proper.Ali [00:09:04]: But, I had expected that something would replace JSON because it's hard to stream JSON ‘cause JSON must be complete and you must have open and close brackets and everything. So it's hard to parse something or validate something while it's being streamed. So people invented all sorts of things that are like, I forget the name of some of these alternatives, but it's something like TOML, something like YAML. But JSON seems to be dominant still.Philip [00:09:30]: The JSON outputs aren't that long, right? Like you could have a long-- ‘cause tool calls also contain the arguments in them and perhaps for a certain tool you might pass like a very long argument. But my impression of the median tool call is that it's a relatively small number of tokens, right? So I would expect that speculators are generally fairly good at something as formatted as JSON. And so you would have like a pretty fast decode step there and that the streaming wouldn't be as valuable, but maybe I'm wrong about that.Ali [00:10:02]: I think you're also bounded by the software or that the model is gonna integrate with if the software is built with JSON for the tool calls or if the company that you'- if your customer says that this is how our software works and our tools are interfaced with JSON, you can ask them to like, change their software and say like, “Yeah, this is gonna be better for the model.” but like with the right training shouldn't be that much of a difference. Also more profitable if it outputs more tokens probably.Swyx [00:10:25]: Depends on your business model.Swyx [00:10:27]: It really depends. But I will say that, as a writer with like experience a lot with generated output, I do try to move from text to JSON text which is very long JSON, right? Like there's paragraphs in every field because I'm trying to structure it, right?Philip [00:10:44]: Right.Swyx [00:10:44]: I want you to first make factual statements, then make opinions then make bullet point summaries, have dates, have entity references have your sources for references, all these things. Anyway, so these are things that like I think people who really experiment with structural output have to really care about. But, let's, let's recurse up the stack a little bit. Before we started recording, you mentioned something really cool, which is that there's a lot of engineering that-- inference engineering that goes on when a new model provider releases a new model, right? So let's call it GLM-5.2, Kimi K3. I had previously assumed, especially if it's like, well, GLM 5 to 5.1 to GLM-5.2, like that you've supported them before. Is it that much work?What It Takes to Support a New Open ModelAli [00:11:26]: It's a lot of work.Swyx [00:11:28]: Yeah. Okay. So like, a lot of people, all you guys, right whenever a new model launch like, people rush to say like, “Oh, Hugging Face supports this, Fireworks supports this, Spacetime supports this,” and I'm like, “Yeah, of course we support it.” But what goes into that? What goes intoPhilip [00:11:40]: I think it's more than just support it too, right? It benefits the consumer a lot. Like I think it was with Kimi K2.5 or GLM-5.2 the latest, there was an inference war, right? X provider is at 90 tokens a second. The next day we're at 150. The nextSwyx [00:11:55]: I kinda kicked that off with the GLM-5.2.Swyx [00:11:58]: I wrote a Twitter article about. It got like half a million views,Ali [00:12:02]: Based on being numberSwyx [00:12:03]: YeahAli [00:12:04]: Or it's for something else.Swyx [00:12:05]: Yeah. Which,Ali [00:12:06]: Oh my GodSwyx [00:12:07]: Which then got everyone really excited about, hey, how can we, bend tracks a little bit further and,Philip [00:12:14]: There's a difference between support the model, as in I can make a token out of this model, and support a model, as in I have a production-ready API from this model.Philip [00:12:26]: Getting to the point of I can make a token out of this model is not that hard because generally the, open source inference engines, vLLM, SGLang of the world oftentimes even receive weights ahead of time, maintainers do, or the people making the model merge PRs to ensure support. So you generally can, just get it working on the standard open source stack without too much pain in most cases. The challenge is, every inference company is gonna have own proprietary stack. Some open source components, some in-house stuff. And for any arbitrary model, there's going to be some new stuff. Sometimes you get lucky, like K, two five to two six was, like, pretty similar.Quantization, Speculators, and Production ReadinessAli [00:13:16]: Yeah. It was pure continued post-trainingPhilip [00:13:18]: YeahAli [00:13:18]: If I remember correctly.Philip [00:13:19]: Even in those cases, there's still stuff you have to do. You have to redo the quantization work. You're taking the model from. Generally, these models are not released in NVFP4, and we want them to be in NVFP4 for maximum Blackwell compatibility. So we have to perform that quantization, and, calibrate the quantization to make sure that we're not causing any regression in the model's intelligence. And then we also have to train the speculator, as we've talked about. Generally, we have. We have ZDR, zero data retention on our model APIs, so we don't know exactly the traffic that people are sending us, but we know what's popular. We know that coding use cases are popular. We know that agents, agentic use cases are popular. So we can get public data sets that are representative of that traffic and train general speculators. Now, with speculators today, you need to train the speculator using the base model itself because you're getting hidden states out of the model from running inference on these specific prompts, and that is the training data you use to create the speculator. So there's that process which you need the real model weights for. And then there's of course just the process of, standing up all the infrastructure behind it, loading all this stuff, testing it. And then when there's a new model with a newer architecture, I think that, like, the DeepSeek models tend to be the most challenging as they have, like, the most novel architectural stuff going on, model after model. But every new model has something. Kimi K2 had. Oh, sorry, GLM-5.2 hadAli [00:14:53]: Sparse attention.Philip [00:14:54]: Yeah,Ali [00:14:54]: YeahPhilip [00:14:54]: the DSA.Ali [00:14:55]: Right. Which is brought from DeepSeek.Philip [00:14:57]: Yeah. AndAli [00:14:59]: So you can copy-paste then?Philip [00:15:01]: It kindAli [00:15:01]: I don't know how this works.Philip [00:15:02]: So, like we had to, like, build support for that into our runtime. And you're right, like it is really interesting the way that all of these open source labs borrow from each other. For example, like GLM-5.2 doesn't have vision. So something that, Haley, a guy on our team, if we could take a look at this, he, like, grafted the Kimi vision encoder onto GLM-5.2.Retrofitting Vision into GLM-5.2Ali [00:15:27]: We'll be training the projector.Philip [00:15:28]: Exactly. So if you think about, like, the encoder, there's the encoder, which is the part that looks at the image and turns it into latent information, and then there's the projector which likeAli [00:15:38]: You can say latent space. It's okay.Philip [00:15:41]: And then there's the projector that maps it onto, the model itself, and then there's the model weights. You don't wanna mess with the model weights because you run a chance of making the model dumber at something else for the purpose of giving it vision. So instead, Haley started with just a projector, which is only a handful of millions of parameters.Ali [00:16:02]: That would be, yeah.Philip [00:16:02]: Yeah.Ali [00:16:03]: Can you show the training one?Ali [00:16:04]: Like the way it groksPhilip [00:16:05]: YeahAli [00:16:06]: Very interesting.Philip [00:16:06]: And maybeAli [00:16:07]: That right therePhilip [00:16:07]: Maybe Ali, you should take it from here. You've got a betterAli [00:16:10]: Ooh, double the sandPhilip [00:16:11]: Understanding of this than I do.Ali [00:16:11]: Yeah. You can see, like, he. The way he trained this is really cool. At the beginning, he was training it using just like, “Here's a picture of a mountain. Can you describe what's in this mountain?” And that caused it just like the first, learning walls. Like here you can see this all we're trying to teach it is to translate the encoded. Like it's already taken the encoder from Kimi K. It's taken the image. It'Philip [00:16:31]: Yeah. FrozenAli [00:16:31]: FrozenPhilip [00:16:32]: With adapter.Ali [00:16:32]: Exactly.Philip [00:16:33]: Yeah.Ali [00:16:33]: So the brain is frozen and the eyes are frozen. It's just we're tryingPhilip [00:16:37]: AlignAli [00:16:38]: Interconnect between the eye and the brain, right? So the projector. And so you take the tokens and then he's like, “Oh, can you describe what's in this image?” And he's like, “Oh, it's a mountain,” or it's a person or it's a human, whatever the case is. But that didn't cause complete understanding. So he changed it such that every image was associated with a data set of questions. Like, does this image have a white male? Does this image have birds in the top corner? Does this image have a scientist in it? All of that stuff. And it would have to answer questions correctly. And using not just training on describing an image, but being able to answer question, another question, answer over time. Like you can see the grokking, which is like genuinely insane, that retrofitting vision into a large LLM can learn to that extent. And even for images that it doesn't perform well on, for instance, if you ask it a picture of like Stephen Hawking, “Who is this?” Maybe it doesn't get it, but it will say something like, “This is Albert Einstein.” Like it still understandsPhilip [00:17:25]: Close enoughAli [00:17:26]: That this is a scientist who is a man who has, some significant achievements, all that stuff. So that's like really cool.Philip [00:17:32]: Yeah. So, we've covered Hao Tian before, who the author of the LLaVA paper that did this, a while ago. And I think that's very foundational work for anyone who hasn't done vision work before.Ali [00:17:41]: Same with the CLIP and MetaCLIP, where you go from just captioning to building out questionsPhilip [00:17:47]: RightAli [00:17:47]: Off the image and how much better you can get performance.Philip [00:17:50]: Right. Right. Right. Yeah. But what's, what's so exciting about this is if you look at a model like this. Now, this is a little bit more of a research project. It's not. It got to 56% on MMLU Pro, I think. So not quite frontier. But if you're running this model, you haven't suffered any loss on your GLM-5.2 quality. If you don't have an image, it'll just behave exactly the way it used to. And ultimatelyAli [00:18:14]: Which in the inference code you literally do not include the other part, right?Philip [00:18:18]: Yeah. You would just skip the encoder if you don't have an image input.Ali [00:18:22]: Okay.Philip [00:18:22]: Just confirming.Philip [00:18:23]: YeahAli [00:18:23]: Does it affect a lot on the overall inference side? Like you're not adding much, you're adding a very small vision encoder. These are typically likePhilip [00:18:30]: They're super fineAli [00:18:31]: Less than a billion parameters, right?Philip [00:18:32]: Yeah. It's, - There's a little bit less standardization among vision encodersSwyx [00:18:37]: YeahPhilip [00:18:37]: So the support matrix can be a little bit, sparser. But overall, yeah, it's a pretty, it's a pretty minor component of the overall system. And ultimately what you get out of the system is all of a sudden you have Kimi Vision, GLM weights, and DeepSeek attention all in one model.Open Source Model Grafting and Franken-MergesPhilip [00:18:56]: And that's, I think, a lot of the power and beauty of open source, is that you can take all of these different components and combine them together into a system that's better than anyoneSwyx [00:19:05]: YeahPhilip [00:19:05]: Can be individually.Swyx [00:19:06]: People used to say that you would also do Franken-merges where you would take likePhilip [00:19:10]: YeahSwyx [00:19:10]: Layers from each model.Swyx [00:19:11]: Does anyone do that anymore?Ali [00:19:13]: Well, to your point previously when you were mentioning like, the work that goes into supporting a model when it first comes out, like GLM-5.2 or MiniMax M3 or whatever the case is. Sometimes you do have to like, you do have to switch out some things. Like, for instance, the MiniMax M3 head uses full attention, and with full attention you end up with this like insane bottleneck in spec dec ‘cause you're doing auto-regressive token generation for three tokens, and you're doing this like N squared over all of the tokens that are in your sequence. Your KV cache is like very large because it's not sparse, it's not top K. So we find it better to like, okay, we're gonna replace this, we're gonna replace this layer with a layer from another model that's using like GQA, for instance. And then just with the right training, you can get it to have the same acceptance rate. So it is very possible to retrofit layers from other models and very much needed. If a layer is like inefficient, the training just becomes the challenge, like how do you ensure that you train it properly? Which again to your earlier point is like the mesh between training and inference. As in like you need very good training in order to do fast inference. That's like, I feel like more and more becoming true.Swyx [00:20:21]: Yeah. Anything else on the support side when you say like get it to fully production ready?Loop Detection, Race Conditions, and Non-DeterminismPhilip [00:20:26]: Yeah. I think that there's also a question of just, we can test a model to a pretty extensive degree, but we're trying to get it out quickly and then you see a bunch of other people test it and you get interesting results. There was an issue with, GLM briefly where we had some like mode collapses where it would just output the same token over and over again for certain prompts on certain temperatures. Like once you expose an endpoint to the real world, there's going to be, so many more varieties of things given to it that you're able to, discover and patch things. So it's not just a, day zero process, it's then like for the first week, for the first month, if a model remains popular, like how do you both fix bugs and then continue to push the envelope on performance?Ali [00:21:21]: What do you mean you don't want your model outputting S?Swyx [00:21:24]: Is there loop detection on that stuff, by the way? It still happens like quite a lot, which is surprising.Ali [00:21:30]: We have like we, in our endpoint, like if a model was to output the same token like four plus times, we just cut the generation. We say like, “Oh, sorry, this-- Like try again,” or like we will reprocess the request. ‘Cause we know then, like if it, like if, yeah, it's four times the same token, it's probably collapsed.Swyx [00:21:45]: Yeah. Is there a way to opt out in case I really want that?Ali [00:21:48]: You want that?Ali [00:21:50]: I think there's a way that we have to handle it. I'm not exactly certain, but I feel like in certain models, like when they output something like you can imagine, like a table for instance, and so they want, they wanna draw like 12 dashes and 12 dashes. Yeah, I think there's a way for that to happen. I think we only do it on certain tokens. Like we exclude certain special characters.Swyx [00:22:07]: Yeah.Ali [00:22:07]: So we only do it on like certain like S is the most common almost. GLM-5.2Swyx [00:22:11]: OhAli [00:22:11]: And I think it was DSV 4 as well. Like you'd just have like looping issues where like you literallySwyx [00:22:17]: ItAli [00:22:17]: Just have like S.Swyx [00:22:18]: Yeah. Is there a special, something special about S? No, just randomlyAli [00:22:21]: It just seems to be the one token involved.Swyx [00:22:23]: Yeah. And it'Philip [00:22:24]: Is thereSwyx [00:22:24]: And it's only temperature 0Ali [00:22:27]: NoSwyx [00:22:27]: Even at other temperaturesAli [00:22:27]: Even at like 0.9 or whatever, it will still, it will still collapse.Swyx [00:22:30]: That's weird, right?Ali [00:22:30]: It's, it is an inference problem to be honest, like a software problem. Like oftentimes, the image you run will-- like NVIDIA will release an image for instance, and if we will upstream the changes from their latest TensorRT-LLM image into our stack, we'll find that it fixes it. Or oftentimes this will only happen in an inference engine that you're using like SGLang. But if you were to switch to vLLM, that isn't the case. So it seems to be like an extremely like deterministic software issue and not really a model issue. It's not like a weights problem. Like I'- we'll say like, “Oh, it's a problem with the quant. We did PTQ wrong,” right? But that isn't, that doesn't make sense because the same weights used with a different inference engine does not repeat the problem. And sometimes it's, the kernels that are being used in the backend have like these very subtle sometimes race conditions, where if you were to use this model hosted on one cluster, you will never get this problem.Swyx [00:23:19]: Oh my God.Ali [00:23:19]: But if you host it on a different cluster, you will. And the reason is the KV cache transfer from a node to node in that one cluster is using a slower interconnect than the node to node in another cluster. So that exposes the race, whereas in another cluster it doesn't. So then you end up just like, okay, this model is not gonna be hosted on this cluster. We're gonna host it on, another cluster because that cluster exposed that problem. But then it ends up with like, okay, is it the software? Is it the model weights or is it the hardware?Swyx [00:23:42]: There is a thing about this with temperature 0 still not being deterministic, right?Ali [00:23:46]: Right.Swyx [00:23:46]: Mostly because of hardware. Even at temperature 0 same model, you won't always get the same output.Swyx [00:23:52]: Even-- But I'm surprised by the race condition one because, I thought PyTorch was a graph that like guarantees that you at least, execute things in the right order.Ali [00:24:02]: Well, yeah, true. Like I'm not, I'm not saying that there is. Like well, you have things like PTL optimizations where like you can start a kernel before the end of the previous kernel, and that's like ‘cause you want to do that because there'sSwyx [00:24:12]: It's like pipeliningAli [00:24:12]: Expense. Exactly.Swyx [00:24:13]: Yeah.Ali [00:24:13]: But it'- But you don't do it cleanly. Like you overlap a little bit of the execution. No, it is very possible that the kernel itself, like that one block that is supposed to be running in this instance of time, that kernel itself has a race condition. For instance, like a missing barrier. Like often if you're designing a kernel and you want it to make it to be very fast, if you don't test it extensively, you'll, you'll have certain threads access data points from registers before they've been written to by other threadsSwyx [00:24:36]: YeahAli [00:24:36]: For example, because like your barrier is wrong or your synchronization was wrong. But yeah, like the testing itself is very difficult in those like, andSwyx [00:24:42]: And there's no like borrow checkerAli [00:24:45]: What does that mean?Swyx [00:24:46]: Like Rust. Like the. If you're trying to have like memory safety It sounds like a comparable problem.Ali [00:24:52]: Well, yes, but you're working in CUDA, right, NVIDIA GPUs. Like- You just need a higher level language like modular Maybe that's what modular is supposed to do. I don't know.Quantization Quality and Vendor FidelityVibhu [00:25:00]: How do you see keeping quality of the model? So you talked about all these steps of, okay, you gotta do quantization, train your own speculative decoderAli [00:25:07]: RightVibhu [00:25:07]: Run on different hardware. Looking at other model providers, okay, you kicked off a inference speed race on the consumer end. What goes into keeping quality the same across them, right? Sure, you can run benchmarksAli [00:25:22]: YeahVibhu [00:25:22]: But, like, how do you determine how much quantization are there standards? What goes intoPhilip [00:25:27]: There's a few things on quality. Most inference optimizations are lossless. KV caching, for example. You are just recomputing or preventing recomputing the same values. Speculation, of course, if a draft token is wrong, it gets rejected. The main lossy optimization is quantization. And that really comes down to, number one, data format, number two, which parts of the model you choose to quantize, which layers, and number three, like doing a lot of calibration on the quantized weights, to ensure that you're preserving all the outliers. There's other tricks that you can do, though. A big one is long context, ‘cause one thing you asked at, right at the beginning is, “Oh, what's gonna happen if I send a 200,000 token request in?” So with a long input sequence, you need to, store a lot more information. You need to process a lot more tokens. And so even if a model has a context of a certain length, you might, as an inference provider, choose to build an API with a shorter context length, and of course a full length one as well. Because if someone doesn't need the full million token context, for example, you can get them better performance. I don't know if that's exactly like quality of the model. The way that I think about quality is to what degree are we faithfully serving the original model? If you think of a golden implementation of a model that performs exactly the way the model is designed to perform, I think of quality as how close are we getting to that, 100% fidelity of the model.Philip [00:27:13]: You can also, of course, think about quality from the training side and how do you push yourself past 100%. But when I think about purely inference optimizations, it's getting faster while staying as close to that 100% fidelity mark as possible. And certainly our standard internally is that, like you should not be able to tell the difference between our API and a, official API. I think Kimi in particular does a good job of vendor benchmarking hereAli [00:27:41]: YesPhilip [00:27:41]: Where they haveAli [00:27:42]: They released an actual vendor benchmark.Philip [00:27:43]: Exactly, yeah.Ali [00:27:44]: ‘Cause they accused, some people, Amazon? There was some provider that was not doing very well on Kimi's benchmark.Philip [00:27:50]: Yeah.Philip [00:27:51]: So, with Reflect we probablyVibhu [00:27:52]: This was a long time ago, right?Philip [00:27:54]: No.Ali [00:27:54]: Yeah, like threeVibhu [00:27:55]: They alsoAli [00:27:55]: Four, five months agoVibhu [00:27:57]: This also happened with, I don't remember which model, but they pulled out quite a few, and then they started a whole chart about this. It might have beenPhilip [00:28:03]: Kimi Vendor Verifier.Ali [00:28:04]: Yeah.Philip [00:28:05]: Yeah.Ali [00:28:05]: Yeah, ‘cause you, ‘cause you'd be pissed, right? Like if you'Philip [00:28:07]: Yeah.Ali [00:28:07]: If like if I'm a consumer and I'm using like Amazon's endpoint for instance, and I've used Kimi and I'm like, “Oh my God, like this is bad,” I'm not gonna say, “Oh, Amazon quantized the model in a bad way.” I'm gonna say, “Oh, Kimi sucks.” Right?Philip [00:28:17]: Yeah.Ali [00:28:17]: So it seems like that makes sense.Philip [00:28:19]: Yeah, they care. They care.Vibhu [00:28:21]: Justifiably.Ali [00:28:21]: Yeah, justifiably.Vibhu [00:28:22]: This is probably a stupid question, but just checking, has anything improved from main quantization?Philip [00:28:28]: Yeah.Vibhu [00:28:28]: Like, is quantization always strictly worse?Ali [00:28:30]: Well technicallyVibhu [00:28:32]: NoAli [00:28:32]: It's a lossy. QuantizationPhilip [00:28:33]: YeahAli [00:28:33]: Is a lossy, it's a lossy implementation.Philip [00:28:36]: Speed improvesVibhu [00:28:36]: Speed improves.Ali [00:28:37]: It the number, likeVibhu [00:28:38]: No, I' always look for inverse scaling laws.Philip [00:28:40]: Yeah.Ali [00:28:40]: Yeah.Vibhu [00:28:40]: This is something I learned from Noam Brown, where like things that normally act in one direction sometimes do.Philip [00:28:45]: Well, technically when you run a benchmark, because these models are deterministic, sometimes your,Ali [00:28:52]: YeahPhilip [00:28:52]: NVFP4 quant is like, two basis points higher than yourAli [00:28:56]: No, it's noise. It's noise.Philip [00:28:57]: Yeah, exactly. I'm like, yeah, it's, it's within. That's why I always say within margin of error.Philip [00:29:01]: And I stopped saying that because everyone assumes that what is, well, within some margin of error, we're barely inside of that to the worst, so we're saying. But yeah, sometimes it's just like, gives you a higher output score. But like Ali said, that's noise. To my knowledge, you're not necessarily making the results better. You're just trying to, again, like keep your fidelity as close to 100% to the original model.Layer Selection, KL Divergence, and Better QuantizationAli [00:29:27]: There is, to your point, research that we did on MP. I don't know if you are able to pullPhilip [00:29:31]: YeahAli [00:29:32]: A tweet we did. One of our research interns, Joshua, I think it's a tweet on how we have 20% better quantized GLM-5.2 than NVIDIA. Essentially what we found throughout like this month research is, okay, quantization is a lossy. It's. You're compressing the data from, occupying 16 bits to occupying, four bits, for instance. And so you're losing some information, and you're trying to minimize that. And so when I say that I'm gonna quantize the model, my job becomes how do I find the layers that I can quantize, and how to find the layers to not. For instance, with image models, I don't quantize modulation layers, and I don't quantize out projections because those two are. Like out projection is what you see as the user. Modulation is what the model sees or understands. Right, exactly. And so to his paper, do you have the. It doesn't have the. Yeah. It's a long paper. I don't know if I can findVibhu [00:30:25]: If there's a part to search or it's probably in the thread.Ali [00:30:28]: It's probably in the thread.Vibhu [00:30:29]: Yeah.Ali [00:30:29]: But the long and the short is it is very possible that quantizing more of the model makes the results. Like if I have a model that I quantize layers one, five, and 10, and another model where I only quantize layers one and It is possible that the model in which I quantized more information is going to perform better because the quantization errors have canceled out. And so what Joshua showed in his mathematical proof where he had like a verifier in, is that you can predict which layers are going to have quantization errors that will cancel out with each other, and you choose to quantize those layers. And so the result of doing this mathematical quantization is you end up with a model that's 20% more quantized than another provider, so you get 20% more throughput of it because there's more layers than running an NVFP4, and your quality is better than that other quant because the layers that you chose to quantize have their errors cancel out, like one layer skewed to the right one layer skewed to the left, one layer skewed to the right. Your final logits distribution is more similar to the original distribution of the model, so you have better fidelity. And so the way we proved this was with KL divergence. So instead of just scoring on the benchmarks, we scored the KL divergence between the logit distribution of the quantized model and the logit distribution of the original full precision model, and we showed that with this technique we get. If your probability distribution on the logits which token it wants to select is more of the same as the original model, you're probably gonna end up staying true to the original model. So yeah, so it seems like previously before this, it seemed like the industry was, well, the more you quantize, the worse it's gonna be, ‘cause the more loss you introduce. That's not exactly, not necessarily true. So yeah, doesn't improve it, but can cancel out.Philip [00:31:57]: I think it might be this, but reminds me a good bit about pruning where you can prune off certain layers.Philip [00:32:03]: But very interesting. Didn't know this was a whole paper you guys put out.Ali [00:32:06]: It's. Fun fact, it was originally 72 pages, this paper, and then we decidedPhilip [00:32:11]: WowAli [00:32:11]: We can't tell. We couldn't release it. So it's now 45.Swyx [00:32:15]: Still 39 pages, so very substantive. We talked about evals and all these things and, like what's possible in terms of speedup? Like it's like probably like the numberInference Speedups and BenchmarkingSwyx [00:32:25]: Thing that people do wanna care about, and it's something that you wrote about in your post. Like official API is 70 tokens per second, and you push it up to 90. Is that like a normal thing?Philip [00:32:36]: So what's cool about working in inference, the reason that I think inference is going to be a useful place to do engineering for a long time, is that if you look at highly optimized domains like, say, finance, if you're in finance, you measure how much better you got in basis points. It's like, “Oh, I got five basis points better, like twentieth of 1% better,” that's huge news because everything is so optimized. When we publish optimizations, it's 20%, it's 100% it's 200%. So there's still probably like a lot further to go, honestly. Like you'll, you'll know that inference is pretty much solved when researchers start publishing about how they got 1% faster at something.Swyx [00:33:19]: Which by the way, because I am from the finance background, in the ‘70s, that was the margin at the time. When you did quantitative finance research, you would findAli [00:33:27]: And like 20%, tens of percent.Swyx [00:33:29]: That's. Yes.Philip [00:33:29]: Yeah.Swyx [00:33:30]: And now it'Philip [00:33:31]: Tiny fractionsSwyx [00:33:32]: For those people interested, look up Andrew Lo's paper. He had a really interesting illustration of quant, stat arb, distribution, narrowing down from like those kinds of 20% differences in the ‘70s, down to nothing today, which is very cool.Philip [00:33:48]: Exactly, and we're at the beginning of the same type of thing. Now benchmarking is hard. I think anyone will tell you that, and benchmarking provider speeds is hard because there's so many variables that go into it. What hardware are you using? How much load do you have on the system? What's the exact nature of the prompts and input and output sequence lengths? All that stuff. But overall, when you start stacking these improvements, you're looking at multiples. You can look at it. The most common form, of course, is TPS, tokens per second, which is bad naming by us in the industry, ‘cause there's two tokens per second. There's tokens per second, the throughput number, and the latency number.Ali [00:34:31]: TTMT, yeah.Philip [00:34:32]: Like total tokens per second out of the, out of the GPU as a throughput number. Most people only care about tokens per second as the latency number, which we should call ITL, intertoken latency, but we don't.Philip [00:34:44]: Anyway, so you can imagine a standard API without many optimizations for a 1 trillion parameter model operating somewhere in the 30 to 50 tokens per second range for reasonable traffic profile. And we generally see the goal of, pushing to 10X that. But, not necessarily day zero, but by stacking enough optimizations, if you have, say like four optimizations, each of which doubles performance. Or sorry, three optimizations, each of which doubles performance, then you stack that up, that's an 8X gain. That's the order of magnitude that we're working with in this space. We're trying to make things substantially faster, not just go from like 70 to 90.Swyx [00:35:38]: Are you saying you've. You have done that?Philip [00:35:40]: So let's say you have as a reasonable baseline, 30 or 40 tokens per second. You can achieve 10X that. So like on GLM-5.2, if you run it unquantized, perhaps on H100s even, and you're just using an off-the-shelf inference engine with no particular optimizations, no speculator, nothing extra around like KV routing, no disaggregation, you're, you're probably, yeah, looking at that like 30 to 40. You think that's like a reasonable baseline?Swyx [00:36:12]: Right. Right.Philip [00:36:12]: To get to something like 10X, there's a lot of trade-offs that you're making. If we're running at more like a 300, 400 tokens per second range, you are using the best hardware possible. You have a optimized speculator. You have done all of your quantization work. You are Seeing a pretty high cache hit rate. You are running with a reasonably small batch size and a parallelism configuration that is tuned for latency versus throughput, but it is possible. So the spreads that you see if you, like, go on artificial analysis or you go on OpenRouter and you look at, the worst provider to the best provider, oftentimes can hit that range. 10X is of course very aggressive. It's oftentimes maybe more of a four to six times improvement. But that's the performance that makes us really excited, is when we can get these huge gains, not just go from 70 to 90 tokens.Stacking Optimizations: NVFP4, Speculation, and DisaggregationAli [00:37:19]: It's also, like, hardware dependent. Like, ifPhilip [00:37:20]: YeahAli [00:37:20]: If you have a thing where you're serving it on just, like, a node of H100s and then you throw, like, you shard the model across, like, four nodes of B200s. Like, you can definitely increase the speed with just throwing more hardware at it. Like, normalizing for the same exact hardware and the same number of GPUs.Philip [00:37:35]: Yeah. Then you're looking at, like, a two to 4X improvementAli [00:37:38]: Right. RightPhilip [00:37:38]: Depending on the inference optimizations. So yeah, it's. Some of it's, what's the call, and some of it's who's the driver.Vibhu [00:37:46]: If you break down the two to 4X, say the example is run GLM-5.2Ali [00:37:51]: YeahVibhu [00:37:51]: On B200sAli [00:37:53]: YeahVibhu [00:37:53]: Single node, right? What's, like, the cost trade-off for effort to get, like, the last bit of juice out versus what should people just think of, right?Ali [00:38:01]: Spectre quantization. Yeah.Vibhu [00:38:03]: Spectre quantization.Ali [00:38:04]: That's, that's, that's like 95%. LikeVibhu [00:38:06]: And how far does that get you? And how easy is that for the average person to do? So say right I wanna throw the weights of GLM-5.2 on a node of B200s, how easy is it to find speculative decoder- decoder model or already quantized model? How much work goes into it?Philip [00:38:23]: If you're doing it up front, it's quite a lot of work. If you're doing it today, there's going to be people who have published things that you can just, you can just grab some NVFP4 weights. You can grab a speculator. Yeah, if we're thinking about, like, what are the 2Xs we're stacking, going from, BF16 to NVFP4 is, it's not quite a 2X, right? It's like. I think it's about, like, 30 to 40%, from 16 to 8, and then another 30 to 40% multiplied from, 8 to 4. So that doesn't quite get you a 2X, but, like, roughly a 2X. Speculator, roughly a 2X. Disagg on top of that if you're able to get enough hardware and put enough traffic through it, another roughly a 2X. And then you add in some, double-digit percent increase from having just a better runtime with, the latest kernels and stuff behind it. And that's how it stacks up.Ali [00:39:21]: YeahPhilip [00:39:21]: So building each of those, like, building the, quantized weights is, for someone who really knows what they're doing, hours to days of work. Building the speculator, again, like, hours to days of work. And the, disagg setup, hours to days. Well okay, but like once you haveAli [00:39:39]: Once set up. Once set up. YeahPhilip [00:39:40]: Yeah, getting disagg working for the first time, I'm saying, of course, is very difficult.Philip [00:39:44]: The marginal implementationAli [00:39:48]: Like, if you're just grabbing, like if you are a person, like just a normal consumer who has access to, like, a node of B200s and you're wondering, “How can I just host it myself?” You don't need to quantize the model yourself. There's always gonna be, like, an open source quantized checkpoint. NVIDIA's gonna push one out if no one else does. You. Usually, the providers will have their own spec dec that they've trained as well. You don't need to train your own spec dec. You can just use that as well.Philip [00:40:09]: Yeah. Like, GLM-5.2 has its own MTP.Ali [00:40:13]: Right. Right.Vibhu [00:40:14]: What's multi token prediction?Philip [00:40:15]: Yes.Ali [00:40:16]: I'm justVibhu [00:40:16]: Can you explain that?Ali [00:40:16]: I'm just an expert.Ali [00:40:18]: I can do it for you in case I get it wrong?Vibhu [00:40:20]: No.Vibhu [00:40:21]: Yeah, you should correct if we're wrong, but their multi-token prediction can be used for self-speculative decoding.Ali [00:40:27]: I'm not sure. I'm not gonna correct that.Vibhu [00:40:28]: Okay. I'm semi-confident in thatAli [00:40:30]: Okay. YeahVibhu [00:40:30]: But someone can check. But it's useful to paint the story of, okay, not just the average person, but say a company wants to switch from serverless inference I wanna throw this up on. I wanna rent some GPUs, throw it up. These are the steps you take to do significantly faster than just put it behind vLLM.Ali [00:40:48]: Right.Vibhu [00:40:49]: I was waiting for a mention of Dynamo.Vibhu [00:40:51]: I feel like, that's supposed to be the baseline that you measure against.Dynamo, KV Routing, and Disaggregation ToolkitsPhilip [00:40:55]: I would think of Dynamo as less of a box system and more of a toolkit for building with. So when we talk about doing aware routing, when we talk about doing KV offloading, when we talk about doing, PD disaggregation, Dynamo fundamentally is. By the way, Dynamo is an open source library from NVIDIA.Ali [00:41:17]: We've done a pod with KylePhilip [00:41:18]: OkayAli [00:41:19]: Kyle Cranin.Philip [00:41:19]: Cool. So then your listeners know then that it supports all the different inference frameworks. And it is multi hardware, which is interesting.Ali [00:41:28]: But it's just a router, it's not like an optimizer layer.Philip [00:41:30]: Yeah. All it does, like, what Dynamo is good at, it is a library for moving information around your cluster, around your hardware. So if you have, KV cache on one place and you need it to be somewhere else, Dynamo coordinates NIXL for you to move that around.Philip [00:41:49]: That doesn't mean that, like, out of the box, you just say, “Pip install Dynamo,” and then you get, like, a massive performance speed up. It's more of a developer toolkit.Ali [00:42:01]: Yeah. I would have said it would. It comes with a set of defaults that you can then swap out.Philip [00:42:06]: It does. If the industry at large, I think, was, like, rolling out all of these deployments, standard, then I think it would be, like, a credible baseline. But, we've got to, we've got to benchmark against, like, what we're seeing in the wild.Speculative Decoding Methods: Medusa, EAGLE, n-Gram, and Spec-SpecVibhu [00:42:23]: I did wanna talk a little bit more about PD disagg, because that is probably, like, number three after quantized and speculative decoding. In your book though, I was just gonna pull out the book.Philip [00:42:31]: Yeah.Vibhu [00:42:32]: Like section 522 on Medusa, 523 on EAGLEPhilip [00:42:35]: YeahVibhu [00:42:36]: 524 on gram.Philip [00:42:37]: It's 55, would be disaggregationAli [00:42:42]: Yeah. Well, no, I just wanted to dwell a little bitPhilip [00:42:44]: YeahAli [00:42:44]: The other. Like, so what do you choose to include? What do you choose to not to include? Because there was all these other techniques.Philip [00:42:51]: Yeah.Ali [00:42:51]: Are these still relevant? Because I think they came out, like, a year and a half ago maybe.Vibhu [00:42:55]: Medusa is quite old.Philip [00:42:56]: Yeah, Medusa's old.Ali [00:42:58]: It was old.Vibhu [00:42:58]: But is it in the book as a good, here'sPhilip [00:43:01]: BaselineVibhu [00:43:01]: Baseline vanilla understand it?Philip [00:43:02]: Like you should know this.Vibhu [00:43:03]: Like I read the paper, I'm like, “ it makes so much sense.”Philip [00:43:05]: Yeah.Philip [00:43:05]: So with the book, I had a couple goals. One was to give people just a working vocabulary for the space as a whole, and the other was to give them some intuition about how each of these techniques works. As I mentioned in my AI Engineer talk, which is the first public addendum to this, the speculation space has moved much faster than everything else. So yeah, even at the time that I wrote the book Medusa, I very much included as a way for people to understand how the space evolved rather than what the most modern technique is. And now of course, there's DFlash, dSpark. There's, there's newer techniques even than EAGLE, although EAGLE is still very commonly used.Ali [00:43:51]: SpecSpecta.Philip [00:43:52]: Yes. Speculative decoding.Vibhu [00:43:54]: What canAli [00:43:56]: Oh, it's a paper by Tri Dao and it's like, it's doing speculative decodingVibhu [00:44:00]: HuhAli [00:44:01]: For the speculative decoder.Philip [00:44:02]: Oh, in spec- oh my God.Ali [00:44:02]: It's literally just an another. It's like, yeah, that's the most simple way to explain it, and it seems like he got trivial speed ups there. But it seems that the complexity with training, it's almost like in our mind at least, it's almost as complex as training GANs. Like it's like a very delicate balance and oftentimes you, it's just but yeah, it's literally speculative decoding on speculative decoding.Vibhu [00:44:21]: Speculative.Ali [00:44:22]: Yeah. We saw this paper.Vibhu [00:44:24]: It's interesting, right?Ali [00:44:24]: Yeah.Vibhu [00:44:24]: I wouldn't even expect it to be very particular to train, I wouldAli [00:44:29]: Right.Vibhu [00:44:29]: The naive part of me is like, okay, train speculative decoder.Ali [00:44:32]: But like, and it makes sense, like the whole idea of speculative decoding is you. It's like, it's like almost like the iPhone auto predict version but for a normal model, right? Like you're just, you're just, generating three tokens and you're like, okay, I'll do prefill on them. And so you save those three turns for your original model. Now your speculative decoder is doing three turns of auto regression, so why not just have an even smaller model?Ali [00:44:53]: The other question there is what are the size of speculators? So say forPhilip [00:44:58]: Right. It's like a billion parameters.Ali [00:45:01]: Like for MiniMax, it's. Yeah. It's like one layer. It's like one 60th of the original model usually.Philip [00:45:06]: Yeah. I think we should do a paper when we get back to the office.Philip [00:45:10]: SpeculativeAli [00:45:11]: SpeculativePhilip [00:45:11]: Decoding.Ali [00:45:13]: No, it's, it does seem like how, when do you stop? But then it also seems like if you're able to train spec-spec decode for instance, right? Like if you're able to have a small model that is accurately predicts what the intermediate speculator is gonna predict, that is able to predict what the original target model's gonna predict, then why not just use that smallest model directly, right?Vibhu [00:45:34]: Yeah. This isAli [00:45:35]: Like it seems likeVibhu [00:45:35]: Adjacent to the routing problem.Ali [00:45:36]: Right.Vibhu [00:45:36]: Yeah.Ali [00:45:36]: Right.Philip [00:45:37]: The thing with speculators is one of the practical constraints on using them is that you do have to run a small model on the same hardware that you're running the big model on. There is a orchestration and resource competition problem inherent in that, and that is one of the constraints on speculation in general, is that draft tokens cost resources to create and cost software complexity to manage. And so if you have like infinitely recursive speculators, you add in quite a bit of that complexity on the actual implementation within the inference engine as well, not just in the training process.Vibhu [00:46:17]: I was gonna say, I would wonder if you could do similar, like distillation and pruning of, it's the same thing, it's just a model. Can we not just distill a lot of the weights, quantize the speculator, out of my domain? The question that also comes up is, this is all for big server workloads, right? How much of this applies to, say I have this MacBook, I wanna run Gemma really efficiently. Similar problems, not the same?Local AI vs. Data Center InferencePhilip [00:46:45]: Pretty different. I talked to Selo, about this on his podcast a couple weeks ago. The difference between inference engineering for the data center and for production workloads versus inference engineering for local AI, is that we start with fundamentally like different constraints and different goals. With local AI, it's how do I fit this model onto my hardware and then make it less dumb? And with data center influence, it's how do I load this model and then make it less slow? And we care about less dumb, and they care about less slow. But the local AI inference engineering ecosystem, I think has a lot for us to learn from in the data center space. They are experts in various forms of quantization, including dynamic quantization that we just don't touch, in the pruning, in the distillation, in the, layer removal. There'Ali [00:47:42]: Layer removal matters less.Philip [00:47:43]: Yeah. There'Ali [00:47:44]: No one loves pruning really.Philip [00:47:45]: Yeah. Well, but the, but they doVibhu [00:47:46]: Which is surprising, right? But that's, that's a whole different thingPhilip [00:47:48]: Just to fit something on the laptop.Ali [00:47:50]: Right.Philip [00:47:50]: So yeah, it's a, it's an interesting, it's an interesting space. Not necessarily that like their techniques make sense for us to do in the data center, because we have different resources and different goals, but more that the process as well as the openness of that field is something to, admire.Ali [00:48:12]: Yeah. Like to your point, like, certain optimizations that would. Like for instance, Turbo Quantum Sharper, like it made such huge hype on that and we did like a whole deep dive on Twitter and like said, what is it? How does it work? Why is it good or not? And it took off and it was implemented on local devices because your memory bandwidth is so slow on like a MacBook, for instance. But try putting the same thing on like an NVIDIA GPU on a B200 Turbo quant would not be. Like, it would not be used. Like, NVIDIA - Like, NVIDIA made it clear that this is not a good optimization, and we've seen it firsthand where the overhead of doing dequantization, quantization of, in the kernel itself with turbo quant kernel, each end is much slower than the time that you save from doing the bandwidth. ‘Cause on the B200s, you have like 3.5 terabytes per second. You don't need decrease the storage that much. You don't need to do, FP4 KV cache. You don't need to use a requant. There's, there's, there's better optimizations to be made. But on Edge devices, it's extremely important, it's extremely useful. So, seems to be, like, different optimizations there, but then they're all uniquely combined with like all you wanna quantize the model, you wanna do speculative decoding, like certain common prefixes with bothPhilip [00:49:18]: Principles.Ali [00:49:19]: Yeah, exactly. Exactly. Exactly.Philip [00:49:20]: They also do a lot of work on, model parallelism, especially over, heterogeneous topology, where you have, some sparks and they are wired together with, Ethernet, DGX sparks.Ali [00:49:35]: Yeah, this is the Exo Labs guys.Philip [00:49:36]: Yeah. You have, a nu

In this month's episode, Ampere explores HBO Max's record-breaking UK rollout, covering strategy, sign-ups, and the streamer bundling landscape. Hosted on Acast. See acast.com/privacy for more information.

Netflix has the largest price gap between ad-free and ad-supported tiers as it also seeks more sports. Ampere forecasts global streaming revenue will exceed $200 billion in 2030.

Ubuntu wants a leaner, stricter GRUB, and your favorite setup may not survive the cut. We break down what's really changing, and the practical ways to adapt. Plus, Chris moves on from one of his favorite open source apps.Sponsored By:Jupiter Party Annual Membership: Put your support on automatic with our annual plan, and get one month of membership for free!Managed Nebula: Meet Managed Nebula from Defined Networking. A decentralized VPN built on the open-source Nebula platform that we love.Support LINUX UnpluggedLinks:

Renault galt lange als Problemkind der europäischen Autoindustrie. Mittlerweile riecht es in Boulogne-Billancourt wieder mehr nach Aufbruch als noch vor ein paar Jahren. In dieser Folge von WAS MICH BEWEGT sprechen wir darüber, wie der französische Autokonzern zwischen Altlasten und den Herausforderungen der Zeit navigiert. Wir schauen auf die jüngere Geschichte und dort vor allem auf die schwierigen Jahre in der Schicksalsgemeinschaft mit Nissan und Mitsubishi – ein Konstrukt, das allerorten bröckelt. Aber es gibt auch eine neue Realität beziehungsweise Perspektive: einen Auto-Riese, der profitabler, disziplinierter und fokussierter ist – und zugleich weiterhin viele Baustellen zu schließen hat. Was bringt die neue Strategie „futuREady“? Wie viel Stärke steckt wirklich hinter Ampere‑Reintegration, globalen Plattformen, Indien‑Offensive und digitalisierter Produktion? Und ist Renault damit stabiler geworden – oder nur beweglicher? Die Details zu Renaults neuer Strategie futREady? https://www.automobil-produktion.de/management/renault-justiert-produktionsstrategie-neu/2624877 Renaults Jahresbilanz: https://www.automobil-produktion.de/news/renault-erwartet-niedrigere-gewinnmargen/2611010 Mehr zu Pascal und Yannick finden Sie auf LinkedIn: Pascal Nagel: https://www.linkedin.com/in/pascal-nagel/ Yannick Tiedemann: www.linkedin.com/in/yannick-tiedemann Hinweis: Die im Podcast getätigten Aussagen spiegeln die Privatmeinung der Gesprächspartner wider und entsprechen nicht zwingend den Darstellungen des jeweiligen Arbeitgebers

CIMA-Wechsel: Wolfgang Lintl übergibt an Krisztian Dolhai. CIMA öffnet sich für STOL-Wettbewerbe und führt klare Trennung von Serial und Open Class bei Motorschirmen ein. Junkers A50 Heritage: DULV-Zulassung für das Nostalgie-UL mit 7-Zylinder-Sternmotor Verner Scarlett 7U (124 PS). Klassische Metallbauweise. Ausgestellt auf der AERO 2026 (Halle A3, Stand 117). Smartflyer SFX-1 Update: Umstellung auf H55-Batteriesystem. Erstflug des hybrid-elektrischen Prototyps auf Herbst 2026 verschoben. Fokus auf CS-23-Zertifizierungsfähigkeit. DFS Luftraum-Knigge: DFS startet Info-Offensive gegen Luftraumverletzungen. Kostenloser Online-Vortrag (11. März, 19 Uhr) und neuer Safety Newsletter mit Lotsen-Perspektiven. In dieser Folge geht es um den Flugzeug Anlasser in der Allgemeinen Luftfahrt und darum, warum dieses Bauteil beim Motorstart oft unterschätzt wird. Wir erklären, wie ein Anlasser im Flugzeug technisch aufgebaut ist, welche Aufgaben er beim Start übernimmt und warum gerade hier Fehlbedienung schnell teuer werden kann. Im Fokus stehen typische Systeme aus der GA Technik, darunter Bendixanlasser und Schub-Schraubtriebanlasser, wie sie bei Lycoming Anlasser- und Continental Anlasser-Systemen eingesetzt werden. Du erfährst, warum Direktanlasser heute in der Allgemeinen Luftfahrt Standard sind und welche Rolle Getriebe, Magnetschalter und Ritzel beim sauberen Einkuppeln spielen. Außerdem geht es um die elektrische Seite: hoher Strombedarf, extreme Belastung und die Frage, warum 100 bis 300 Ampere beim Motorstart Flugzeug absolut normal sein können. Die Folge zeigt, weshalb langes Orgeln den Anlasser beschädigen kann und warum Überhitzung zu den häufigsten Ursachen für Ausfälle gehört. Gerade bei Kälte, Startproblemen oder einem Hot Start Flugzeug machen viele Piloten die gleichen Fehler. Wir besprechen, wie man Startprobleme systematisch einordnet, wann Priming sinnvoll ist und warum Vorwärmen oft günstiger ist als eine Reparatur. Auch die Unterschiede zwischen älteren und neueren Anlassersystemen bei Continental- und Lycoming-Motoren werden verständlich erklärt. Dazu kommt ein klarer Blick auf typische Schwachstellen wie Relais, Kabel, Kontaktprobleme und innere Defekte des Anlassers. Die Folge richtet sich an Privatpiloten, Technikinteressierte und alle, die Flugzeugtechnik nicht nur bedienen, sondern wirklich verstehen wollen. Wer wissen will, wie man einen Flugzeug Anlasser schont, typische Fehler vermeidet und teure Schäden in der Allgemeinen Luftfahrt verhindert, bekommt hier eine kompakte und praxisnahe Einordnung. Feuerlöscher-Regelung: EASA gibt nach massivem IAOPA-Einsatz nach. Keine Zertifizierung für nicht-fest installierte (leicht entnehmbare) Feuerlöscher unter 1200 kg MTOM im NCO-Betrieb. Empfehlung: Kompakte Aerosol-Löscher.

【节目简介】2022年，雷诺CEO Luca De Meo把雷诺一拆五并押宝Ampere，专注软件和电驱动2024年，安培公司IPO失败2025年，Luca宣布离开雷诺加入奢侈品开云Kering集团2026年，新CEO福兰把安培关了本期《孤岛车谈》新雨和两位战略专家聊聊这个曾经以“欧洲首家电动汽车与软件纯玩家”自居的明珠企业的过往，以及雷诺的未来。安培和雷诺的职能差异【话题成员】罗新雨 底盘电子系统工程师王涵 德国某车企的战略规划Eric 合资中国分部的战略部门剪辑 PSC，猫又，罗新雨片尾曲 Si j'allais en Chine by Les Enfantastiques (2013)【时刻文稿】17:42 安培公司的目标21:04 安培公司的车型22:10 twingo的意义:21个月搞定37:23 雷诺和宝马、大众的对比48:00 雷诺和中国车企的对比51:09 小蓝灯的故事【参考链接】【【专属】乘用车底盘系统开发 车辆动力学原理应用与正向开发工程实践 吴旭亭 系统构建车身动力学底盘知识体系书籍】#小程序://机械工业出版社旗舰店/商品/I4N8mLuPmjWkmRt【官网 车用动力电池系统设计与制造 中国汽车工程学会 电芯产品设计 电池系统产品设计 动力电池产品设计制造方法技术书籍】#小程序://机械工业出版社旗舰店/商品/P8isKji8jO5DkNc【汽车创新：前沿技术背后的科技原理】#小程序://机械工业出版社旗舰店/商品/7tltQzCQfJUWRVi【官网 广义车规级电子元器件可靠性设计与开发实践 左成钢 系统介绍汽车电子零部件的可靠性设计与开发 汽车电子 汽车工业技术书籍】#小程序://机械工业出版社旗舰店/商品/dBujAN68sEk1Rzl【智能驾驶：产品设计与评价】#小程序://机械工业出版社旗舰店/商品/Q8KWriuNDGdzlSs【官网 智能底盘关键技术及应用 线控执行 融合控制 失效运行 张俊智 智能底盘核心线控执行系统关键技术书籍】#小程序://机械工业出版社旗舰店/商品/5R5ZjdGhScib14A2024年安培的IPO夭折（英文，2024）：https://media.renaultgroup.com/renault-group-decides-to-cancel-amperes-ipo-and-confirms-its-ev-and-software-strategy/?lang=fra据消息人士透露，雷诺将关闭安培（Ampere）项目部门，该项目是德梅奥时代的旗舰项目（英文，2026）：https://www.reuters.com/business/renault-shut-down-ampere-unit-flagship-project-de-meo-era-sources-say-2026-01-21/

Ampere traverse une période de transition, avec sa réintégration progressive au sein du groupe Renault. Mais pour le constructeur français, pas question de lever le pied sur l'électrique. Au contraire. La dernière annonce en date le montre : la recherche continue, et elle vise un point névralgique de la voiture électrique, la batterie.Renault vient en effet d'officialiser un partenariat entre Ampere, sa filiale dédiée aux véhicules électriques, et l'entreprise espagnole Basquevolt. Objectif : accélérer le développement des batteries dites « lithium métal », une technologie considérée comme l'une des pistes les plus prometteuses pour dépasser les limites actuelles.Aujourd'hui, la plupart des voitures électriques utilisent des batteries lithium-ion à électrolyte liquide. L'électrolyte, c'est le milieu dans lequel circulent les ions lithium entre les deux électrodes de la batterie lors de la charge et de la décharge. Dans la technologie développée par Basquevolt, cet électrolyte liquide est remplacé par un électrolyte polymère solide, associé à une anode avancée en lithium métal.Ce changement n'est pas anodin. Selon les partenaires, il permettrait un « saut technologique majeur » en matière de densité énergétique — autrement dit, plus d'énergie stockée pour un même volume ou un même poids. À la clé : des batteries plus compactes, plus légères, dotées d'une meilleure stabilité thermique — un enjeu crucial pour la sécurité — et capables de se recharger plus rapidement. On parle ici de la grande famille des batteries solides, sur lesquelles plusieurs constructeurs, notamment asiatiques, travaillent activement. Basquevolt affiche d'ailleurs une ambition claire : devenir un leader européen dans ce domaine stratégique.Mais au-delà des performances, l'enjeu est aussi industriel. Le recours à un électrolyte polymère permettrait, selon le communiqué, de simplifier le processus de fabrication. Résultat annoncé : environ 30 % d'investissement en moins par gigawattheure dans une gigafactory classique, et 30 % d'énergie économisée par kilowattheure produit. Pour Renault, l'équation est simple : gagner en autonomie, en sécurité et en compétitivité. Hébergé par Acast. Visitez acast.com/privacy pour plus d'informations.

It's EV News Briefly for Thursday 26 February 2026, everything you need to know in less than 5 minutes if you haven't got time for the full show.Patreon supporters fund this show, get the episodes ad free, as soon as they're ready and are part of the EV News Daily Community. You can be like them by clicking here: https://www.patreon.com/EVNewsDaily EV LIST PRICES FALL AS GAS GUZZLER PRICES RISENew EV list prices (excluding Tesla) dropped 2.3%, or roughly $1,500, from an average of $63,327 in September 2025 to $61,860 in January 2026, while average new gas-powered vehicle prices rose 2.5% to $47,427 over the same period. The sharpest cuts came after the federal EV tax credit expired, with the Hyundai IONIQ 5 leading the slide at a 13.8% drop of over $7,000, followed by the Chevrolet Equinox EV at nearly $4,000 off — six models in total posted drops above 5%. FORD TEASES EUROPE CAR RETURN AFTER FIESTA, FOCUSFord CEO Jim Farley used the Q4 2025 earnings call to signal "exciting plans" for passenger cars in Europe, framing the comeback as a selective, profitable return to specific segments rather than a volume land grab. Two new EVs built on Renault's Ampere platform are expected in the subcompact segment from the Ford–Renault partnership, with new passenger cars set to start arriving in 2027 under a new dedicated Europe passenger-car leadership role. UBER EXPANDS EV RIDES ACROSS EIGHT UK CITIESUber has rolled out its EV ride option to eight more UK cities — Birmingham, Manchester, Leeds, Sheffield, Edinburgh, Cambridge, Belfast and Merseyside — at standard UberX prices, after falling short of its pledge to run an all-electric London fleet by end-2025. Only 40% of London miles are now covered by EVs, with UK General Manager Andrew Brem citing charging access as "the biggest barrier," prompting Uber to announce driver support measures including discounted home and public charging in partnership with Pod Point. BMW TALKS PRICE FLOOR TO DODGE EU MINI DUTYBMW and the European Commission are in advanced talks to replace the EU's 20.7% countervailing duty on China-made Mini BEVs with a minimum import price agreement, according to Handelsblatt — covering the Mini Cooper Electric and Mini Aceman, both built at BMW's Zhangjiagang joint venture with Great Wall Motor. The approach would mirror the "price undertaking" the EU accepted from Volkswagen Anhui in early February, which freed the Cupra Tavascan from countervailing duties in exchange for a confidential price floor, volume cap and EU investment commitments. EU CITY BUS SALES HIT 60% ZERO-EMISSIONSix in ten new city buses registered across the EU in 2025 were zero-emission — 56% battery-electric and 4% fuel cell — a dramatic jump from just 12% when the Clean Vehicles Directive was adopted in 2019. Five member states hit 100% zero-emission city bus sales in 2025 (Bulgaria, Denmark, Estonia, Latvia and Slovenia), and Transport & Environment says a fully zero-emission EU city bus market is achievable as early as 2028.​MG2 SET FOR 2027 UK LAUNCH AT £20,000MG will enter the electric supermini segment in 2027 with the all-new MG2, targeting a starting price of around £20,000 (~$25,200), to take on rivals including the Renault 5, Citroën e-C3, Fiat Grande Panda and the incoming VW ID. Polo. The car will use the newer E3 architecture from the MG4 Urban, run front-wheel drive with a torsion-beam rear axle for cost efficiency, and feature a 12.8-inch touchscreen with physical climate controls — a reveal is expected in the second half of 2026. MG CONFIRMS MGS9 PHEV SEVEN-SEATER FOR UKMG will launch the MGS9 plug-in hybrid SUV in the UK later in 2026 as its new flagship, offering three full adult-sized rows and targeting rivals such as the Peugeot 5008, Kia Sorento and Skoda Kodiaq at a value-led price point. The model already holds a five-star Euro NCAP rating and could reach UK showrooms as early as summer 2026, extending MG's line-up to 11 models. AUSTRALIA NVES DATA SHOWS HYBRIDS DO THE HEAVY LIFTAustralia's National Vehicle Emissions Standard published its first half-year performance data (July–December 2025), showing EVs made up roughly 12% of new vehicles supplied, with about two-thirds of manufacturers — including BYD and Polestar — meeting their fleet-wide emissions targets. Petrol- and hybrid-focused brands such as Mazda and Hyundai fell short and face penalties if they don't improve, while the data reveals that near-term emissions gains are leaning more on efficient hybrids than on full EVs. LECTRON ADAPTERS WIN UL 2252 SAFETY CERTIFICATIONLectron has earned UL 2252 safety certification across its full range of EV charging adapters — covering J3400, CCS1 and J1772 in both AC and DC variants — with its two DC adapters handling up to 500 amps at 1,000 volts for peak power of 500 kW, and built-in thermal sensors that trigger derating if heat rises during fast charging. The certification comes as the North American charging landscape remains split between NACS and CCS1 on DC networks and J1772 on AC infrastructure, making a certified bridging adapter an increasingly essential tool for EV drivers navigating the transition.

- Hyundai Surpasses GM Market Cap - VW Overtakes Tesla in Europe - Chinese EVs Surging in Mexico - Data Center Chip Shortages Affecting OEMs - New Mercedes S-Class and Level 4 Tech - VW Project Trinity and Recycling Hub - Renault Bringing Ampere In-House - Mazda and Volvo Hybrid Pivot on EVs - Nissan Dealer Profit Crisis

- Hyundai Surpasses GM Market Cap - VW Overtakes Tesla in Europe - Chinese EVs Surging in Mexico - Data Center Chip Shortages Affecting OEMs - New Mercedes S-Class and Level 4 Tech - VW Project Trinity and Recycling Hub - Renault Bringing Ampere In-House - Mazda and Volvo Hybrid Pivot on EVs - Nissan Dealer Profit Crisis

Security, compliance, and resilience are the cornerstones of trust.   In this episode, Lois Houston and Nikita Abraham continue their conversation with David Mills and Tijo Thomas, exploring how Oracle Cloud Infrastructure empowers organizations to protect data, stay compliant, and scale with confidence.   Real-world examples from Zoom, KDDI, 8x8, and Uber highlight these capabilities.   Cloud Business Jumpstart: https://mylearn.oracle.com/ou/course/cloud-business-jumpstart/152957 Oracle University Learning Community: https://education.oracle.com/ou-community LinkedIn: https://www.linkedin.com/showcase/oracle-university/ X: https://x.com/Oracle_Edu   Special thanks to Arijit Ghosh, David Wright, Kris-Ann Nansen, Radhika Banka, and the OU Studio Team for helping us create this episode.     -------------------------------------------------------------   Episode Transcript:   00:00 Welcome to the Oracle University Podcast, the first stop on your cloud journey. During this series of informative podcasts, we'll bring you foundational training on the most popular Oracle technologies. Let's get started! 00:26 Lois: Hello and welcome to the Oracle University Podcast! I'm Lois Houston, Director of Communications and Adoption with Customer Success Services, and with me is Nikita Abraham, Team Lead: Editorial Services with Oracle University. Nikita: Hi everyone! In our last episode, we started the conversation around the real business value of Oracle Cloud Infrastructure and how it helps organizations create impact at scale. Lois: Today, we're taking a closer look at what keeps the value strong — things like security, compliance, and the technology that helps businesses stay resilient. To walk us through it, we have our experts from Oracle University, David Mills, Senior Principal PaaS Instructor, and Tijo Thomas, Principal OCI Instructor.  01:12 Nikita: Hi David and Tijo! It's great to have you both here! Tijo, let's start with you. How does Oracle Cloud Infrastructure help organizations stay secure? Tijo: OCI uses a security first approach to protect customer workloads. This is done with implementing a Zero Trust Model. A Zero Trust security model use frequent user authentication and authorization to protect assets while continuously monitoring for potential breaches. This would assume that no users, no devices, no applications are universally trusted. Continuous verification is always required. Access is granted only based on the context of request, the level of trust, and the sensitivity of that asset. There are three strategic pillars that Oracle security first approach is built on. The first one is being automated. With automation, the business doesn't have to rely on any manual work to stay secure. Threat detection, patching, and compliance checks, all these happen automatically. And that reduces human errors and also saving time. Security in OCI is always turned on. Encryption is automatic. Identity checks are continuous. Security is not an afterthought in OCI. It is incorporated into every single layer. Now, while we talk about Oracle's security first approach, remember security is a shared responsibility, and what that means while Oracle handles the data center, the hardware, the infrastructure, software, consumers are responsible for securing their apps, configurations and the data. 03:06 Lois: Tijo, let's discuss this with an example. Imagine an online store called MuShop. They're a fast-growing business selling cat products. Can you walk us through how a business like this can enhance its end-to-end security and compliance with OCI? Tijo: First of all, focusing on securing web servers. These servers host the web portal where customers would browse, they log in, and place their orders. So these web servers are a prime target for attackers. To protect these entry points, MuShop deployed a service called OCI Web Application Firewall. On top of that, the MuShop business have also used OCI security list and network security groups that will control their traffic flow. As when the businesses grow, new users such as developers, operations, finance, staff would all need to be onboarded. OCI identity services is used to assign roles, for example, giving developers access to only the dev instances, and finance would access just the billing dashboards. MuShop also require MFA multi-factor authentication, and that use both password and a time-based authentication code to verify their identities. Talking about some of the critical customer data like emails, addresses, and the payment info, this data is stored in databases and storage. Using OCI Vault, the data is encrypted with customer managed keys. Oracle Data Safe is another service, and that is used to audit who has got access to sensitive tables, and also mask real customer data in non-production environments. 04:59 Nikita: Once those systems are in place, how can MuShop use OCI tools to detect and respond to threats quickly? Tijo: For that, MuShop used a service called OCI Cloud Guard. Think of it like a security operation center, and which is built right into OCI. It monitors the entire OCI environment continuously, and it can track identity activities, storage settings, network configurations and much more. If it finds something risky, like a publicly exposed object storage bucket, or maybe a user having a broad access to that environment, it raises a security finding. And better yet, it can automatically respond. So if someone creates a resource outside of their policy, OCI Cloud Guard can disable it.  05:48 Lois: And what about preventing misconfigurations? How does OCI make that easier while keeping operations secure?  Tijo: OCI Security Zone is another service and that is used to enforce security postures in OCI. The goody zones help you to avoid any accidental misconfigurations. For example, in a security zone, you can choose users not to create a storage bucket that is publicly accessible. To stay ahead of vulnerabilities, MuShop runs OCI vulnerability scanning. They have scheduled to scan weekly to capture any outdated libraries or misconfigurations. OCI Security Advisor is another service that is used to flag any unused open ports and with recommending stronger access rules. MuShop needed more than just security. They also had to be compliant. OCI's compliance certifications have helped them to meet data privacy and security regulations across different regions and industries. There are additional services like OCI audit logs for traceability that help them pass internal and external audits. 07:11 Oracle University is proud to announce three brand new courses that will help your teams unlock the power of Redwood—the next generation design system. Redwood enhances the user experience, boosts efficiency, and ensures consistency across Oracle Fusion Cloud Applications. Whether you're a functional lead, configuration consultant, administrator, developer, or IT support analyst, these courses will introduce you to the Redwood philosophy and its business impact. They'll also teach you how to use Visual Builder Studio to personalize and extend your Fusion environment. Get started today by visiting mylearn.oracle.com.  07:52 Nikita: Welcome back! We know that OCI treats security as a continuous design principle: automated, always on, and built right into the platform. David, do you have a real-world example of a company that needed to scale rapidly and was able to do so successfully with OCI? David: In late 2019, Zoom averaged 10 million meeting participants a day. By April 2020, well that number surged to over 300 million as video conferencing became essential for schools, businesses, and families around the world due to the global pandemic. To meet that explosive demand, Zoom chose OCI not just for performance, but for the ability to scale fast. In just nine hours, OCI engineers helped Zoom move from deployment to live production, handling hundreds of thousands of concurrent meetings immediately. Within weeks, they were supporting millions. And Zoom didn't just scale, they sustained it. With OCI's next-gen architecture, Zoom avoided the performance bottlenecks common in legacy clouds. They used OCI functions and cloud native services to scale workloads flexibly and securely. Today, Zoom transfers more than seven petabytes of data per day through Oracle Cloud. That's enough bandwidth to stream HD video continuously for 93 years. And they do it while maintaining high availability, low latency, and enterprise grade security. As articulated by their CEO Erik Yuan, Zoom didn't just meet the moment, they redefined it with OCI behind the scenes. 09:45 Nikita: That's an incredible story about scale and agility. Do you have more examples of companies that turned to OCI to solve complex data or integration challenges? David: Telecom giant KDDI with over 64 million subscribers, faced a growing data dilemma. Data was everywhere. Survey results, system logs, behavioral analytics, but it was scattered across thousands of sources. Different tools for different tasks created silos, delays, and rising costs. KDDI needed a single platform to connect it all, and they chose Oracle. They replaced their legacy data systems with a modern data platform built on OCI and Autonomous Database. Now they can analyze behavior, improve service planning, and make faster, smarter decisions without the data chaos. But KDDI didn't stop there. They built a 300 terabyte data lake and connected all their systems-- custom on-prem apps, SaaS providers like Salesforce, and even multi-cloud infrastructure. Thanks to Oracle Integration and pre-built adapters, everything works together in real-time, even across clouds. AWS, Azure, and OCI now operate in harmony. The results? Reduced operational costs, faster development cycles, governance and API access improved across the board. KDDI can now analyze customer behavior to improve services like where to expand their 5G network.  Next up, 8 by 8 powers communication for over 55,000 companies and 160 countries with more than 3 million users, depending on its voice, video, and messaging tools every day. To maintain that scale, they needed a cloud that could deliver low latency global availability and high performance without blowing up costs. Well, they moved their video meeting services from Amazon to OCI and went live in just four days. The results? 25% increase in performance per node, 80% reduction in network egress costs, and a significantly lower overall infrastructure spend. But this wasn't just a lift and shift. 8 by 8 also replaced legacy tools with Oracle Logging Analytics, giving their teams a single view across apps, infrastructure, and regions. 8 by 8 scaled up fast. They migrated core voice services, deployed over 300 microservices using OCI Kubernetes, and now run over 1,700 nodes across 26 global OCI regions. In addition, OCI's Ampere-based virtual machines gave them a major boost, sustaining 80% CPU utilization and more than 30% increased performance per core and with no degradation. And with OCI's Observability and Management platform, they gained real-time visibility into application health across both on-prem and cloud. Bottom line, 8x8 represents yet another excellent example of a company leveraging OCI for maximum business results. 13:24 Lois: Uber handles more than a million trips per hour, and Oracle Cloud Infrastructure is an integral part of making that possible. Can you walk us through how OCI supports Uber's needs? David: Uber, the world's largest on-demand mobility platform, handles over 1 million trips every hour. And behind the scenes, OCI is helping to make that possible. In 2023, Uber began migrating thousands of microservices, data platforms, and AI models to OCI. Why? Because OCI provides the automation, flexibility, and infrastructure scale needed to support Uber's explosive growth. Today, Uber uses OCI Compute to handle massive trips serving traffic and OCI Object Storage to replace one of the largest Hadoop-based data environments in the industry. They needed global reach and multi-cloud compatibility, and OCI delivered. But it's not just scale, it's intelligence. Uber runs dozens of AI models on OCI to support real-time predictions up 14 million per second. From ride pricing to traffic patterns, this AI layer powers every trip behind the scenes. And by shifting stateless workloads to OCI Ampere ARM Compute servers, Uber reduced cost while increasing CPU efficiency. For AI inferencing, Uber uses OCI's AI infrastructure to strike the perfect balance between speed, throughput, and cost. So the next time you use your Uber app to schedule a ride, consider what happens behind the scenes with OCI. 15:18 Lois: That's so impressive! Thank you, David, for those wonderful stories, and Tijo for all of your insights. Whether you're in strategy, finance, or transformation, we hope you're walking away with a clearer view of the business value OCI can bring. Nikita: Yeah, and if you want to learn more about the topics we discussed today, visit mylearn.oracle.com and search for the Cloud Business Jumpstart course. Until next time, this is Nikita Abraham… Lois: And Lois Houston signing off! 15:48 That's all for this episode of the Oracle University Podcast. If you enjoyed listening, please click Subscribe to get all the latest episodes. We'd also love it if you would take a moment to rate and review us on your podcast app. See you again on the next episode of the Oracle University Podcast.

Eis knackt. Thorsten geht trotzdem. Klopsköpfe auf Brandenburger Seen, Shih Tzus in Fahnen, Fische unter Hunden – oder waren es Hunde unter Fischen? BERLIN FRIERT. Ein Keller. Zehn Tage. Null Grad. Ein Mann, der alles trägt, was er besitzt, und trotzdem zittert wie ein Werbetauscher ohne Gas. Brenner kaputt! Brenner kaputt! Brenner kaputt! – schreit das Universum synchron in Schulen, Gewerbehallen, Zehlendorfer Vulkangruppen. Verschwörung? Vielleicht. Wer Geheimdienste podcastet, sieht überall Geheimdienste. Logisch. Zwei plus zwei plus drei gleich DUNKEL. Die Sicherung heißt fünfzig, aber was nützt das, wenn die Wand braune Striche spricht? Elektrotechnik für Fortgeschrittene, erklärt von Männern mit trockener Haut und kurzer Zündschnur. Trump beendet Kriege (acht! seiner Meinung nach!). Venezuela hat Schweröl. Aladdin El-Mafaalani hat Hoffnung. Und Guse hat seit zehn Tagen nicht geduscht. Wie soll man duschen, wenn's unter Null ist? Wie? – fragt die winterliche Hackfresse ins Mikrofon. Fußboden bei Totze: drin. Fotos bei Guse: nicht. Vorwürfe: ja. Fischgräten: vielleicht. HOFFNUNG ist das Ende. Hoffnung und: Bis später, wa?

In this episode, we interview Jeff Wittich, Chief Product Officer at Ampere. Learn about his journey up the hardware stack from Process Engineer to Product Officer, how the Innovator's Dilemma led to the founding of Ampere, and how this will impact the next generation of AI.

In this final episode of the year, Ampere explores who is winning the Christmas catalogue across broadcast TV, streaming, and sport. How are UK broadcasters, global streamers, and live sports rights holders competing for festive audiences, and what do changing Christmas viewing habits mean for content strategy in an increasingly fragmented market? Hosted on Acast. See acast.com/privacy for more information.

Oracle is increasing capital spending by $15 Billion, but they didn't raise their revenue guidance. In this video, we break down the multiple factors impacting Oracle's (ORCL) Fiscal Q2 2026 earnings. Debt is rising toward $110 Billion, free cash flow has swung to negative $10 Billion, and a one-time sale of Ampere is masking the true net income.We analyze why Oracle is selling its chip unit to SoftBank, why "Remaining Performance Obligations" (RPO) are not set in stone revenue, and why we are tax-loss harvesting our position until the cash flow improves.Join us on Discord with Semiconductor Insider, sign up on our website: www.chipstockinvestor.com/membershipSupercharge your analysis with AI! Get 15% of your membership with our special link here: https://fiscal.ai/csi/Sign Up For Our Newsletter: https://mailchi.mp/b1228c12f284/sign-up-landing-page-short-formIf you found this video useful, please make sure to like and subscribe!*********************************************************Affiliate links that are sprinkled in throughout this video. If something catches your eye and you decide to buy it, we might earn a little coffee money. Thanks for helping us (Kasey) fuel our caffeine addiction!Content in this video is for general information or entertainment only and is not specific or individual investment advice. Forecasts and information presented may not develop as predicted and there is no guarantee any strategies presented will be successful. All investing involves risk, and you could lose some or all of your principal.Chapters:00:00 - Oracle Earnings: Why the Narrative Soured 01:25 - The Guidance Miss: $15B More Spend, Zero Extra Revenue 02:18 - RPO Explained: Is the $523 Billion Backlog Real? 03:55 - CapEx vs. Revenue: Comparing Oracle to Other Hyperscalers 05:10 - Cash Flow Alert: The Swing to Negative $10 Billion 06:20 - The Ampere Sale: Why Larry Ellison Sold His Chip Unit 07:30 - Strategy Shift: Moving to "Chip Neutrality" 08:24 - The Debt Load: Total Debt Approaching $110 Billion 08:45 - Final Verdict: Why We Are Selling for Tax Loss Harvesting#Oracle #StockAnalysis #ORCL #CashFlow #Investing #AIInfrastructure #TechStocks #BalanceSheetNick and Kasey own shares of Oracle

- Volvo Throws Luminar Under the Bus - Elkann Proposes CO2 Emission Easing - Xiaomi's EVs Now Profitable, Stock Takes a Beating - Nissan Adds a Mitsubishi to Its Lineup - The Think Behind Ford's Dealership Design - Ford to Sell CPO Cars on Amazon - Next-Gen Ford Bronco and Ranger - Trafic Van is Renault's 1st SDV - Ram Offers 'Free-Agent' Rides in NASCAR Truck

- Volvo Throws Luminar Under the Bus - Elkann Proposes CO2 Emission Easing - Xiaomi's EVs Now Profitable, Stock Takes a Beating - Nissan Adds a Mitsubishi to Its Lineup - The Think Behind Ford's Dealership Design - Ford to Sell CPO Cars on Amazon - Next-Gen Ford Bronco and Ranger - Trafic Van is Renault's 1st SDV - Ram Offers 'Free-Agent' Rides in NASCAR Truck

NEW CTP for IBIT Under/Over - Looks like the OVER Presidential PARDONS and A King's EVICTION All Excited - Making headway on the Deficit ! PLUS we are now on Spotify and Amazon Music/Podcasts! Click HERE for Show Notes and Links DHUnplugged is now streaming live - with listener chat. Click on link on the right sidebar. Love the Show? Then how about a Donation? Follow John C. Dvorak on Twitter Follow Andrew Horowitz on Twitter   Warm-Up - NEW CTP for IBIT - Under/Over - Looks like the OVER - Presidential PARDONS and A King's EVICTION - All Excited - Making headway on the Deficit ! Markets - DJIA hits new ATH - after big moves last week - Buyers stepped in (again) - NASDAQ - lagging as AI trade is questioned - THEREFORE - what is happening is a simple rotation again - ALL IN! Just back from a wedding in NJ - Did some Apple picking, hot cider and donuts! Pardon Me... - President Donald Trump has pardoned a long list of his political allies for their support or involvement in plans to overturn the 2020 presidential election, according to the Department of Justice's Pardon Attorney, Ed Martin. - Several others (20+) were pardoned as well including some that plead guilty. Over the Pond - King Charles III on Thursday stripped his disgraced brother Prince Andrew of his remaining titles and evicted him from his royal residence after weeks of pressure to act over his relationship with sex offender Jeffrey Epstein, Buckingham Palace said. - After the king's rare move, which follows years of shameful scandals, he will be known as Andrew Mountbatten Windsor and not as a prince, and he will have to vacate his Royal Lodge mansion near Windsor Castle. Debt - The U.S. government's gross national debt surpassed $38 trillion Wednesday, a record number that highlights the accelerating accumulation of debt on America's balance sheet. - It's also the fastest accumulation of a trillion dollars in debt outside of the COVID-19 pandemic — the U.S. hit $37 trillion in gross national debt in August this year. - “During his first eight months in office, President Trump has reduced the deficit by $350 billion compared to the same period in 2024 by cutting spending and boosting revenue,” White House spokesman Kush Desai said in a statement, adding that the administration would pursue robust economic growth, lower inflation, tariff revenue, lower borrowing costs and cuts to waste, fraud and abuse. - Petersen Foundation: “Along with increasing debt, you get higher interest costs, which are now the fastest growing part of the budget,” Peterson added. “We spent $4 trillion on interest over the last decade, but will spend $14 trillion in the next ten years. Interest costs crowd out important public and private investments in our future, harming the economy for every American.” - Debt Growing by $69,000 per second over the the past year... Stock New - BIG Softbank - Softbank sells entire stake in NVDA - $5.83 Billion - To soften the blow, they said that it as because they are using it to redeploy further into AI - OpenAi to be specific - Also sold part of T-Moblie and using margin loan against ARM to fund the $22.5 BILLION investment in OpenAI - “This should not be seen, in our view, as a cautious or negative stance on Nvidia, but rather in the context of SoftBank needing at least $30.5bn of capital for investments in the Oct-Dec quarter, including $22.5bn for OpenAI and $6.5bn for Ampere,” Rolf Bulk, equity research analyst at New Street Research, told CNBC. - IMPORTANT SO MARKETS DO NOT GET SPOOKED: ?[SoftBank] made a point of saying that it wasn't any view on NVIDIA. ... At the end of the day, they are using the money to invest in other AI related companies,” he said. Coreweave Earnings - The provider of infrastructure for artificial intelligence companies, reported better-than-expected third-quarter revenue on Monday, but the company delivered disappointing full-year...

It's EV News Briefly for Thursday 09 October 2025, everything you need to know in less than 5 minutes if you haven't got time for the full show. Patreon supporters fund this show, get the episodes ad free, as soon as they're ready and are part of the EV News Daily Community. You can be like them by clicking here: https://www.patreon.com/EVNewsDaily MERCEDES EV SALES RETURN TO GROWTH https://evne.ws/4ogQtTX NEW INVESTMENT TO HELP MUNRO EV SCALE M‑SERIES PRODUCTION https://evne.ws/48mukyO NEW TESLA MODEL Y STANDARD TRIM MAY COME TO THE UK https://evne.ws/42SNysn HOW TESLA MODEL Y MATCHES UP AGAINST FORD MUSTANG MACH‑E https://evne.ws/4o36xIK NISSAN OFFERS ROGUE-BASED E-POWER TO RIVALS https://evne.ws/42w801W WEAVEGRID AND LG VENTURES PARTNER ON GRID-INTERACTIVE EVS https://evne.ws/4oaqa1l VOLVO Q3 SALES SEE RISING EV SHARE AMID FALLING DELIVERIES  https://evne.ws/42y7Irk EV INTEREST MAY STAY STABLE AFTER TAX CREDIT END https://evne.ws/4oa0Sk8 LEAPMOTOR B10 TO LAUNCH IN UK  https://evne.ws/47bTQ8N GERMANY'S GREEN LEADER CRITICISES MERZ ON EV POLICY https://evne.ws/46ZCLxu INSTAVOLT UK IS UPGRADING ITS CHARGERS  https://evne.ws/4o6q1Mr CROSS-PAVEMENT EV CHARGING NOW ALLOWED IN NORTHERN IRELAND https://evne.ws/4mR6sqt AMPERE CONFIRMS 40% EV COST CUT TARGET https://evne.ws/48hPyOh BMW LAUNCHES ELECTRIC STANDUP PADDLE BOARD https://evne.ws/4pZqQs7 Mercedes-Benz EV sales return to growth   Mercedes-Benz achieved a 9% year-on-year increase in Q3 2025 electric sales after six straight quarters of decline, thanks to strong CLA demand. Electric vans also surged nearly 96% year-over-year, positioning Mercedes for further EV momentum with new models coming next year.   Munro EV scales M‑Series production   Scottish automaker Munro EV is expanding its Glasgow plant with new funding to boost M‑Series all-terrain EV production and create up to 300 jobs. The investment supports plans to build up to 5,000 rugged electric vehicles annually and strengthen Scotland's return to vehicle manufacturing.   Tesla Model Y Standard trim for UK   Tesla's new Standard trim Model Y, recently launched in the US, could reach the UK by 2026 offering improved efficiency and a lower entry price. Starting near £44,000, the new version provides a smooth, efficient design update while keeping the range close to 365 miles WLTP.   Nissan e‑Power supply plans   Nissan plans to supply its Rogue-based hybrid using e‑Power technology to partners like Ford and Stellantis to expand electrified options. The move could boost production at its Tennessee plant and give other automakers an affordable route to add hybrid powertrains.   WeaveGrid and LG Ventures partnership   LG Ventures has invested in WeaveGrid to develop smarter, grid‑interactive EV charging solutions. The partnership combines LG's battery technology with WeaveGrid's AI software to optimize energy use and strengthen grid reliability as EV adoption grows.   Volvo Q3 EV share rises  Volvo's U.S. EV deliveries jumped 71% year-over-year in Q3 2025 even as total sales dipped 9%, driven by the EX30 and EX90 models. Electrified vehicles now make up a growing share of Volvo's lineup, underscoring its steady shift to full electrification.   EV demand remains solid post‑credit   A J.D. Power report finds U.S. EV interest holding steady even as federal tax credits phase out, suggesting stable long-term demand. Consumers increasingly cite improved choice, charging access and quality as the main reasons for considering electric vehicles.  Leapmotor B10 UK launch  Leapmotor debuts the B10 electric crossover in the UK from under £30,000, boosted by a manufacturer “Leap Grant.” With 270 miles of range, rapid 168 kW charging, and premium equipment, the B10 extends affordable EV options across 53 UK dealerships.   Greens criticise Merz EV stance   German Green leader Katharina Dröge criticised Chancellor Merz's call to abandon the EU's 2035 combustion engine ban, calling it damaging to industry progress. She urged stronger EV incentives, more charging infrastructure, and social leasing to keep Germany competitive in the global transition.   InstaVolt upgrading UK chargers   InstaVolt has begun nationwide upgrades replacing older 50 kW units with faster, more accessible 120 kW chargers. The new design shortens charge times and improves convenience, backed by renewable power and near‑perfect network reliability.   Cross‑pavement charging in Northern Ireland   Northern Ireland now allows residents without driveways to install cross‑pavement home charging channels for EVs. The scheme offers much cheaper charging than public points, helping hundreds of thousands access affordable home charging for the first time.   Ampere confirms 40% cost‑cut target   Renault's EV unit Ampere reaffirmed its goal to reduce EV costs by 40% by 2028 while reaching break-even in 2025. Success would strengthen Renault's competitiveness and make its next generation of electric models more affordable.   BMW electric paddle board  BMW and SipaBoards have launched an electric self‑inflating stand‑up paddle board featuring modular batteries and digital controls. Priced at €3,990, the innovative board combines quiet motorization, safety features, and app connectivity for water sports enthusiasts.

In this episode, Ampere explores how gaming, sport, and media are evolving around young audiences. From social discovery to shifting engagement, we ask: what do consumers really want next? Hosted on Acast. See acast.com/privacy for more information.

- Overcapacity Is Killing China's Auto Industry - Renault To Build EVs For Nissan and Mitsubishi - Nissan Consolidates Global Design Operations - Toyota Creates App to Get PHEV Owners to Charge More - BMW Adopts New Sales Model in Europe - U.S. BEV Registrations Up 6% In July - Rivan Breaks Ground on Georgia Plant - Ford Cuts 1K Jobs at German EV Plant  - Racing for The Visually Impaired

- Overcapacity Is Killing China's Auto Industry - Renault To Build EVs For Nissan and Mitsubishi - Nissan Consolidates Global Design Operations - Toyota Creates App to Get PHEV Owners to Charge More - BMW Adopts New Sales Model in Europe - U.S. BEV Registrations Up 6% In July - Rivan Breaks Ground on Georgia Plant - Ford Cuts 1K Jobs at German EV Plant  - Racing for The Visually Impaired

In this episode, Ampere speaks with Mark Endemaño, Managing Director at AlixPartners, about the rise of broadcaster and streamer partnerships. From Disney+ and ITV in the UK to Netflix and TF1 in France, we explore what has triggered this new wave of collaboration, why it matters now, and how it could reshape the future of the media landscape. Hosted on Acast. See acast.com/privacy for more information.

Anne Currie is joined by Sean Varley, Chief Evangelist and VP of Business Development at Ampere Computing, a leader in building energy-efficient, cloud-native processors. They unpack the energy demands of AI, why power caps and utilization matter more than raw compute, and how to rethink metrics like performance-per-rack for a greener digital future. Sean also discusses Ampere's role in the AI Platform Alliance, the company's partnership with Rakuten, and how infrastructure choices impact the climate trajectory of AI.

vGPU Unlock for Ampere and Ada??? And I Bought An EV!!! 

We are LIVE on hashtag #JSATV from DCD Connect New York with Sean Varley of Ampere! Power constraints are a growing challenge in digital infrastructure, and Ampere Computing is leading the way with energy-efficient, ARM-based processors that cut power use by 30-60%.Sean shares how Ampere is shaping the future of AI and compute efficiency —plus, insights on their planned acquisition by SoftBank.

En el Radar Empresarial de hoy analizamos la salida de Luca de Meo de Renault. El italiano deja el grupo francés después de cinco años. Lo deja de forma voluntaria y desempeñará el cargo hasta el próximo 15 de julio. De Meo asegura que “busca nuevos retos fuera del sector de la automoción". Y es que ante todo, el empresario siempre ha sido un amante del motor, con experiencia en marcas como Toyota, Fiat o Seat. Solo hace un mes presentó su libro Diccionario sentimental del automóvil. Esos nuevos retos podrían llevarle hasta Kering, según Le Figaro. La marca de lujo francesa, que engloba firmas de renombre como Gucci, Yves Saint Laurent o Balenciaga se ha fijado en el buen hacer del ejecutivo en Renault, donde cogió las riendas en un momento muy pero que muy difícil. El 1 de julio de 2020 asumió el cargo de CEO de Renault. El fabricante automovilístico registró una pérdida neta de más de 8 mil millones de euros ese mismo año. Sus ingresos cayeron entonces casi la mitad. Fue entonces cuando De Meo se puso manos a la obra y presentó el plan “Renaulution”: Resurrección, Renovación y Revolución. En la primera fase la marca pretendía hacer un plan de recortes ambicioso en tres años: ya en 2022, la compañía logró reducir estas pérdidas en más de 2.000 millones de euros. En las fases de Renovación y Revolución la marca iba a apostar por la innovación tecnológica, la reestructuración de la marca y la incorporación de nuevos modelos. En noviembre de 2022 la marca crea las marcas Ampere y Power para desarrollar hardware y así reducir su dependencia de China. Además, lanza al mercado los modelos Twingo eléctrico de 20.000 euros y el Renault 5 E-Tech. Otra de las misiones que tenía el ejecutivo italiano era cambiar la alianza con Nissan. De las múltiples tensiones que muchas veces amenazaron con romper dicha alianza se pasó a una relación más calmada. De hecho, la salida de De Meo ha provocado que la marca japonesa vaya a reducir su participación en Renault, como ha asegurado su CEO, Iván Espinosa. Hasta aquí llega la influencia de De Meo, que ahora tiene la difícil tarea de devolver los días de gloria a Kering, después de que sus acciones hayan perdido el 80% de su valor desde 2021. Sabemos que De Meo se va a Kering pero ¿quién le sustituirá en Renault? Parece que el mercado mira a dos muy posibles: Wayne Griffiths, que dejó el cargo de director ejecutivo de Seat en marzo y que ya negoció con el presidente de Stellantis, John Elkann, para ser el CEO del grupo y Josep Maria Recasens, actual CEO de Ampere, filial de coches eléctricos de Renault, división que le ha dado grandes resultados al fabricante automovilístico.

In today's Daily Fix:Grand Theft Auto 6's delay disappointed a lot of gamers, but probably disappointed retailers even more. Analysts at Ampere are suggesting that the delay will cost the games industry around $2.7 billion dollars. The money isn't gone, however, merely moved to next year when GTA 6 drops on May 26. Speaking of that date, Take-Two CEO Strauss Zelnick is confident that Rockstar Games can hit that date. Let's hope so—there's $2.7 billion riding on it. And finally, Nintendo will make a return to the Gamescom show in Germany this year. Nintendo skipped last year's show, but with a new console and a handful of brand new first-party games coming this year, they have plenty to show off.

A new week means new questions! Hope you have fun with these!Which actress played Talia al Ghul in the 2012 film The Dark Knight Rises?From 1884-1915, a German colony known as German South West Africa was forcibly developed in what present-day African country?Don't consider it a weakness if you don't know, but what is the longest word in English containing only one vowel?Between Volt, Amp (or Ampere), and Watt, which is the unit of electric potential?In 2011, which football club became the first non-English club to participate in the Premier League?You'd have to have a heart of glass not to like Chicago's best-loved food.In Hindu mythology, what is the name of the elephant-headed God of New Beginnings, Wisdom and Luck?Who was the longest reigning heavyweight boxing champion?Which country currently uses the koruna as currency?MusicHot Swing, Fast Talkin, Bass Walker, Dances and Dames, Ambush by Kevin MacLeod (incompetech.com)Licensed under Creative Commons: By Attribution 3.0 http://creativecommons.org/licenses/by/3.0/Don't forget to follow us on social media:Patreon – patreon.com/quizbang – Please consider supporting us on Patreon. Check out our fun extras for patrons and help us keep this podcast going. We appreciate any level of support!Website – quizbangpod.com Check out our website, it will have all the links for social media that you need and while you're there, why not go to the contact us page and submit a question!Facebook – @quizbangpodcast – we post episode links and silly lego pictures to go with our trivia questions. Enjoy the silly picture and give your best guess, we will respond to your answer the next day to give everyone a chance to guess.Instagram – Quiz Quiz Bang Bang (quizquizbangbang), we post silly lego pictures to go with our trivia questions. Enjoy the silly picture and give your best guess, we will respond to your answer the next day to give everyone a chance to guess.Twitter – @quizbangpod We want to start a fun community for our fellow trivia lovers. If you hear/think of a fun or challenging trivia question, post it to our twitter feed and we will repost it so everyone can take a stab it. Come for the trivia – stay for the trivia.Ko-Fi – ko-fi.com/quizbangpod – Keep that sweet caffeine running through our body with a Ko-Fi, power us through a late night of fact checking and editing!

LOTUS CUTS MORE JOBSFollowing the announcement in November last year, Lotus are cutting a further 270 jobs in the UK. The proposal is claimed to be to “enhance competitiveness”. The jobs will be lost from the Hethel production facility and not the other two sites in the UK. If you want to learn more, click this Autocar article link here. ASTON MARTIN RESTRUCTURING FINANCES Lance Stroll is looking to increase his stake in Aston Martin, taking it to 35%. At the same time, the company is selling shares in the F1 team, to help raise further cash. You can read more, by clicking this Yahoo! Finance article here. STELLANTIS ENDS LEAPMOTOR PRODUCTION IN POLANDStellantis has suddenly and without explanation, ended production of the Leapmotor T03 in Poland. The company had already backtracked on opening a second Polish production facility, back in November 2024. If you wish to find out more, click this electrive article link here. AMPERE GETS NEW CEOLuca de Meo has handed over the running of Ampere to Josep Maria Recasens, who was the Chief Operations Officer. Ampere is the electric vehicle and software development spin off from Renault. de Meo was CEO of Ampere whilst being CEO of the Groupe Renault. Click this electrive article link to read more. LUCID WINS NIKOLA FACTORY AUCTION The bankruptcy auction of Nikola's assets has enabled Lucid Motors to buy the production facility in Arizona. You can find out more, by clicking this TechCrunch article here. 6000 NEW PUBLIC CHARGERS FOR BRIGHTONChar.gy and Brighton and Hove city have penned a deal that means 6,000 new public on-street chargers will be installed. Char.gy takes advantage of utilising the greenest/cheapest grid production, with different costs depending on when you charge. You can read more, by clicking this EV Powered article here. GM AND CRUISE SUED BY INVESTORSInvestors have sued GM and Cruise, claiming that the companies overstated the capability and state of their autonomous project. The lawsuit contends that GM also asserted that the AV division was key to the financial performance of the company, with the company claiming $50 billion profits were expected by 2030. If you wish to find out more, click this GM Authority link here. HCVA LAUNCHES CAREERS HUBThe Historic and Classics Vehicle Alliance (HCVA), has launched a careers hub that will allow for a central point that will highlight job opportunities in the industry. To learn more,

NEW CAR REGISTRATION FIGURES MARCH 2025The March 2025 new car registration figures have been revealed by the Society of Motor Manufacturers and Traders (SMMT). They show an increase that you have to go back to 2019, to find bettered. Electric vehicles, unsurprisingly, increased as buyers tried to avoid the new VED and luxury car tax. Private, on the one hand, had a surprising jump but then not when thinking about the tax changes. This is expected to be a bit of a one off month, with the numbers next month being a lot lower. Click this link to read more, from SMMT's own article. UK GOVERNMENT TINKERS WITH ZEV The UK Government has announced changes to the requirements of the Zero Emission Vehicle mandate. These add up to tinkering. Finally, we get a decision on hybrid vehicles post 2030, where they will still be sold. Small manufacturers will be exempt from the requirements. All OEMs will be given more flexibility in achieving the required target rates, with the CO2 credit transfer system remaining in place until 2029. In reality though, as we continue to be in a cost of living crisis and vehicles are too expensive, trying to mandate a market will remain difficult. You can read more, by clicking this Autocar article link here. CAR BRANDS AND TRADE BODY FINED BY CMATen car companies and the SMMT have been fined by the Competition and Markets Authority (CMA) for agreeing to not compete with each other when advertising the amount of recycling that could be achieved with their cars. Additionally, the second charge was for anticompetitive practices when it came to recycling end of life vehicles. Following an admission to this, they are to pay £77 million in fines. In Europe a similar investigation has found 15 brands guilty and they will pay £382.7 million, as a consequence. To find out more, click this Autocar link here. MCLAREN MERGES WITH FORSEVEN TO SECURE FUTUREMcLaren has merged with Forseven, to create McLaren Group Holdings, thereby securing its immediate future. Both entities are owned by CYVN Holdings, which is the Abu Dhabi sovereign wealth fund. Forseven has a raft of vehicles almost ready for production but needed a badge to stick on them. Click the EVO article link here to read more. RENAULT AND NISSAN RESET PARTNERSHIP Renault and Nissan have renegotiated their partnership, following a few tumultuous years. Nissan will no longer contribute to the Ampere project, will rebadge the Twingo city car vehicle and have sold their joint-venture stake in India to Renault. If you would like to learn more, click this electrive article link here. If you like what we do, on this show, and think it is worth a £1.00, please consider supporting us via Patreon. Here is the link to that CLICK HERE TO SUPPORT THE PODCAST NEW NEW CAR NEWS - Nissan MicraNissan has revealed the first images of the new Micra, which is a rebadged Renault 5 EV. Once again the company revealed no details or specifications....

Welcome and Episode Overview- Will Townsend hosts episode 220 of G2 on 5G- Anshel Sag joins as fellow analyst- Discussion of latest 5G and 6G developmentsVerizon's AI Assistant for Small Businesses- Launch of Verizon Business Assistant for customer interactions- Text-based chatbot to automate and improve engagement- Potential for retention and customer stickinessSateliot's Funding and Satellite Technology- €15 million grant from Spanish government- Compact "briefcase satellites" for IoT applications- Comparison with other satellite communication companiesNvidia's AI RAN and 6G Ambitions- Partnership with T-Mobile and Cisco for AI-native approach- Nvidia's influence on potential 6G standards- Aggressive positioning in telecom AI developmentT-Mobile's Spectrum Deal- Agreement with Grain Management LLC- Swap of 800 MHz for 600 MHz spectrum and cash- Strategic move to bolster T-Mobile's coverageSoftBank's AI and Telecom Initiatives- Development of Large Telecom Model (LTM)- Acquisition of Ampere Computing- Positioning for AI and telecom infrastructure growthClosing Thoughts and Contact Information- Invitation for listener feedback on 5G topics

This week in Portland startup news, Ampere eyes a big exit, GeekWire considers an award for Jama's big exit, and the shuttering of 18F creates a positive tech movement in Portland. Let's get into it…PORTLAND STARTUP LINKS- Ampere acquired by SoftBank https://amperecomputing.com/press/softbank-group-to-acquire-ampere-computing- Vote for Portland's Jama in the GeekWire Awards https://www.surveymonkey.com/r/2025AwardsVoting- TiE Oregon Westside Pitch https://www.tieoregon.org/pitch-oregon/westside-pitch- 18F https://en.wikipedia.org/wiki/18F- Portland Digital Corps https://digitalcorpspdx.org/- Vancouver Washington startup incubator https://www.youtube.com/watch?v=B5R4PY7lRWU- Goodbye Genie, Hello Rides https://www.linkedin.com/pulse/we-built-ai-companion-kids-entertained-thousands-every-monsef-8kg1c/PORTLAND OREGON STARTUP NEWS00:00 Portland Oregon startup news00:15 Build Week with Vanport Studio and Instrument01:45 Ampere Computing acquired for $6.5 billion05:50 Jama exit up for GeekWire Awards09:45 TiE Oregon Westside Pitch12:15 18F shuttering creates Portland win18:20 A startup incubator in Vancouver Washington19:50 Darius Monsef launches new startupFIND RICK TUROCZY ON THE INTERNET AT…- https://patreon.com/turoczy- https://linkedin.com/in/turoczy- https://bsky.app/profile/turoczy.bsky.social- https://siliconflorist.substack.com/ABOUT SILICON FLORIST ----------For nearly two decades, Rick Turoczy has published Silicon Florist, a blog, newsletter, and podcast that covers entrepreneurs, founders, startups, entrepreneurship, tech, news, and events in the Portland, Oregon, startup community. Whether you're an aspiring entrepreneur, a startup or tech enthusiast, or simply intrigued by Portland's startup culture, Silicon Florist is your go-to source for the latest news, events, jobs, and opportunities in Portland Oregon's flourishing tech and startup scene. Join us in exploring the innovative world of startups in Portland, where creativity and collaboration meet.ABOUT RICK TUROCZY ----------Rick Turoczy has been working in, on, and around the Portland, Oregon, startup community for nearly 30 years. He has been recognized as one of the “OG”s of startup ecosystem building by the Kauffman Foundation. And he has been humbled by any number of opportunities to speak on stages from SXSW to INBOUND and from Kobe, Japan, to Muscat, Oman, including an opportunity to share his views on community building on the TEDxPortland stage (https://www.youtube.com/watch?v=Cj98mr_wUA0). All because of a blog. Weird.https://siliconflorist.com#pdx #portland #oregon #startup #entrepreneur

US equity futures are higher after Wednesday's gains. European markets opened little changed, and Asian equities ended mixed. Markets reacted positively to a dovish-leaning Fed policy decision. The Fed kept rates unchanged and signaled greater uncertainty around the economic outlook, driven largely by ongoing tariff concerns. Although policymakers downgraded the US GDP growth outlook and raised inflation forecasts, Chair Powell indicated that tariff-driven inflation pressures may be temporary. The Fed also decided to slow the pace of quantitative tightening starting April 1st. Elsewhere, Ukraine's President Zelenskiy voiced support for a ceasefire focused on energy infrastructure during discussions with President Trump, with both sides agreeing to collaborate on air defense systems and Ukraine's electricity supply.Companies Mentioned: SoftBank, Ampere, Carlyle, Oracle, Beacon Roofing Supply, QXO, NVIDIA, Intel

Listen for the latest from Bloomberg News See omnystudio.com/listener for privacy information.

Apple's reported shakeup among its AI leadership has intensified the spotlight on the ongoing AI talent wars among mega-cap tech companies. And Softbank's acquisition of Ampere deepens CEO Masayoshi Son's bet on data centers. It's a strategic move in one of the hottest battlegrounds in tech: AI infrastructure. 

ARM SoCs (system on a chip) have become a hot topic in the computing world in the past few years. Apple branded ‘Apple Silicon', Qualcomm's Snapdragon, Ampere's Altra, and others have been disruptive in a world once dominated by x86/x64-based systems from Intel & AMD.In Part 1 of a two-part series on ARM in postproduction, we explore some of the essentials of ARM systems, including:Basics of ARM vs x86/x64 processorsRISK vs Non-RISK CPUsThe flexibility & scalability of ARMThe goal of a uniform product architecture and its advantage for a company like AppleGPU design/performance - the surprise of Apple's ARM implementationThe appeal and benefits of efficiency and low power consumptionThe benefits of unified memoryPackage scalability - faster/more cores, multiple SoCsDoes clock speed matter with ARM SOCs?Additional benefits - onboard encode/decode abilitiesAre SoC GPUs ever going to be on par with discrete GPUs? Will discrete GPUs ever come to Apple ARM systems?In part two, we'll dive a bit deeper, exploring additional topics, including how cloud-based ARM computing could be a game changer for cost-effective, decentralized post workflows, what the future may hold for workstations from Apple and others, and much more. 

Jeff Wittich Ampere is making waves in the telecom industry by unveiling its Ampere® Altra family of processors, engineered to deliver high performance while slashing power consumption—a dual imperative for next-generation 5G, edge computing, and AI workloads. In a press release released this morning, the company detailed its ambitious push into the telecom market, and during a live interview with Technology Reseller News' publisher Doug Green, Chief Product Officer Jeff Wittich shed further light on the innovation behind the new offering. A Cloud-Originated Innovation Now Heading to Telecom Historically recognized as a cloud and CPU company, Ampere has spent the past seven years perfecting processors that can do more with less. “We started in the cloud, where power constraints became a major challenge with rising workloads like AI,” explained Wittich during the podcast. “The problems we solved in the cloud—delivering high performance with low power—are the same issues facing telecom operators today.” Ampere's approach diverges sharply from conventional methods that simply add power to boost performance. Instead, the Ampere Altra family was designed from the ground up to optimize energy efficiency without sacrificing computing muscle. This unique balance is increasingly critical as telecom networks evolve to support not only traditional voice and data but also a burgeoning array of AI-driven applications. The Shift to ARM and the Ecosystem Play One of the standout features of Ampere's new processors is their ARM-based architecture. As industries shift away from x86 solutions—a trend already evident in smartphones, edge devices, and cloud servers—the telecom market is now poised to benefit from the enhanced efficiency and scalability that ARM-based designs offer. The press release highlighted several key partnerships that underpin Ampere's strategy. Collaborations with industry players such as Parallel Wireless, SUSE, Canonical, SynaXG, Fujitsu, and Supermicro are setting the stage for robust, scalable telecom solutions. Parallel Wireless, for example, is leveraging Ampere's processors in its GreenRAN™ platform—a hardware-agnostic, cloud-native radio access network solution that supports 2G, 4G, and 5G technologies. “By combining our energy-efficient processors with innovative partners, we're able to tap into new customer segments like telco carriers,” Wittich noted. “This collaboration not only optimizes performance but also significantly reduces operational costs and environmental impact.” Lower Power, Higher Performance—and a Greener Future At the heart of this breakthrough is Ampere's innovative CPU design. As Wittich emphasized, “The CPU really is the star here. That is really where the innovation is coming from.” This chip not only delivers several times more tokens per second for AI workloads but also operates within an impressively low power envelope—sometimes as low as 40 watts. Such efficiency means these processors can run without active cooling, significantly reducing both operational costs and the environmental impact associated with high-powered data centers. A Timely Announcement As the telecom industry braces for rapid expansion in 5G and AI-driven services, Ampere's announcement arrives at a critical juncture. With its comprehensive ecosystem of hardware and software partners, the company is poised to lead the transformation toward more energy-efficient, high-performing network infrastructures. For those looking to dive deeper into Ampere's solutions, additional details are available on the company's website, amperecomputing.com, where information spans from cloud applications to cutting-edge telecom stacks. With both the press release and this live discussion underscoring its innovative approach, Ampere is set to make a significant impact on the telecom landscape—delivering the power, efficiency, and sustainability that modern networks demand. Visit www.

- Big AI land grab - Nuclear power for data centers - New European-origin quantum system in Spain - Softbank eyes Ampere [audio mp3="https://orionx.net/wp-content/uploads/2025/02/HPCNB_20250210.mp3"][/audio] The post HPC News Bytes – 20250210 appeared first on OrionX.net.

We have new info regarding the future of RADEON, and discuss Nvidia Blackwell! [SPON: Support MLID by downloading Filmora 14 for FREE: https://bit.ly/4avRYHV ] [SPON: Use "brokensilicon“ at CDKeyOffer for $23 Win11 Pro: https://www.cdkeyoffer.com/cko/Moore11 ] 0:00 AI thinks Dan is Steve & Jensen's Jacket looks CHEAP (Intro Banter) 5:09 Does Intel really want to minimize Battlemage Production? (Corrections) 9:30 Nvidia RTX 5090 Reviews – Why it's Underwhelming 16:10 How isn't the 5090 better at Raster, RT, OR AI?! 27:16 Nvidia DLSS 4 Multi-Framegen Reviewed 40:07 RTX 5090 & 5080 Launch sounds like “another Ampere” 44:47 Some version of FSR 4 is planned for RDNA 3/3.5 49:09 RX 9000 Performance Leaked…and Delayed 52:37 (NEW Leak) Lisa Su KILLED Big RDNA 4 1:06:21 Intel Battlemage B570 10GB Reviewed & “Launched” 1:13:57 Nintendo Switch 2 (Kinda) Revealed 1:22:51 PCIe 7.0, Melting Connectors Solved, 4-Slot Titan, Strix Halo Performance (Wrap-Up) 1:30:30 AMD CPU Codenmaes, 9950X3D V-Cache Benefits (Final RM) https://www.anandtech.com/show/21425/intel-lunar-lake-architecture-deep-dive-lion-cove-xe2-and-npu4 https://en.wikipedia.org/wiki/Arrow_Lake_(microprocessor) https://www.techspot.com/review/2944-nvidia-geforce-rtx-5090/ Ridiculous: https://www.techspot.com/review/2944-nvidia-geforce-rtx-5090/#2025-01-23-image-2-jpg LTT underwhelmed by AI Performance: https://youtu.be/Q82tQJyJwgk?si=Gy7rr2EL99AGxtaJ&t=1047 Capped Power Scaling: https://x.com/ComputerBase/status/1882433556678615247 Mostly Underwhelming 8K Performance: https://www.techradar.com/computing/gaming-pcs/nvidia-rtx-5090-8k-performance-has-blown-me-away-already-and-its-mainly-thanks-to-multi-frame-generation https://www.storagereview.com/review/nvidia-geforce-rtx-5090-review-pushing-boundaries-with-ai-acceleration MSI 5090 Suprim water-cooled review: https://www.techpowerup.com/review/msi-geforce-rtx-5090-suprim-liquid/ https://youtu.be/Q82tQJyJwgk?si=WeiVlNImWXI0Qc6f&t=788 https://www.youtube.com/watch?v=B_fGlVqKs1k&ab_channel=HardwareUnboxed https://youtu.be/bZ6NeSGad4I?si=YaSbxKUOzhuDfmy4 https://youtu.be/bZ6NeSGad4I https://youtu.be/0Zk688QKrAg https://www.youtube.com/watch?v=YuGlXL3uKKQ&ab_channel=Moore%27sLawIsDead https://x.com/McAfeeDavid_AMD/status/1881435903358628047 AMD Marketing forgot to delete a scheduled Tweet: https://videocardz.com/newz/somebody-didnt-get-a-memo-amd-ad-claims-you-can-enjoy-gaming-on-radeon-rx-9070-xt-already Seems like it might even be late March: https://videocardz.com/newz/amd-radeon-rx-9070-preorders-to-start-on-march-23-according-to-major-us-retailer https://www.techspot.com/review/2943-intel-arc-b570/ https://www.tomshardware.com/pc-components/gpus/intel-arc-b570-review-asrock-challenger-oc-tested https://youtu.be/itpcsQQvgAQ?si=GV4q8534eqmSTBi0 https://www.forbes.com/sites/jasonevangelho/2025/01/16/3-crucial-takeaways-from-the-switch-2-reveal-trailer/ https://www.techspot.com/news/106409-pcie-70-spec-nearing-completion-promising-16gbs-lane.html https://quasarzone.com/bbs/qn_report/views/444251 https://youtu.be/RDr1pr_c6ts?si=vup3QhjEZ0eOCaQj https://www.techspot.com/news/106369-microsoft-considered-shutting-down-xbox-before-major-game.html https://cdn.wccftech.com/wp-content/uploads/2025/01/AMD-Ryzen-AI-Max-395-Strix-Halo-APU-With-Radeon-8060S-iGPU-_-Gaming-Benchmarks.png?_gl=1*1524w9j*_ga*MTA0ODI5MjA1MS4xNzI3MzAxNzI4*_ga_591JRXV2QC*MTczNzkyNDQ5Ny40My4xLjE3Mzc5MjQ0OTguMC4wLjA

In this episode, host Brian Fuller interviews Jeff Wittich, Chief Product Officer at Ampere Computing. Ampere, founded in 2018, has revolutionized data center computing with its ARM-based processors. Together, they dive into the challenges of modern computing, the role of generative AI, and the company's strategies to redefine performance and efficiency in data centers and edge environments.

- U.S. New Car Sales Up 11% In October - BYD Sets Sales Record - Leapmotor Sales Leap 110% - Porsche, Audi Profits Plummet - Scout Could Build Vehicles for Other Automakers - Ford Turns to Lectron for NACS - Walmart Chooses GM Brightdrop Vans - Carvana Is Screaming Ahead - Ampere Targets 40% EV Cost Cuts - Toyota Reveals Open-Air Land Cruiser Concept - Infiniti Creates Luxurious Tailgating Concept

- U.S. New Car Sales Up 11% In October - BYD Sets Sales Record - Leapmotor Sales Leap 110% - Porsche, Audi Profits Plummet - Scout Could Build Vehicles for Other Automakers - Ford Turns to Lectron for NACS - Walmart Chooses GM Brightdrop Vans - Carvana Is Screaming Ahead - Ampere Targets 40% EV Cost Cuts - Toyota Reveals Open-Air Land Cruiser Concept - Infiniti Creates Luxurious Tailgating Concept

Listen to the latest System76 updates including Halloween sales on laptops and desktops, the release of the second alpha of Cosmic and the introduction of the Thelio Astra, an arm-based desktop catering to autonomous vehicle and software-defined vehicle developers. Interviews with Joe Speed from Ampere and System76's Mechanical Engineer Britain and Product Manager Tony explore the technical specifications, engineering excellence, and collaborative efforts behind the Thelio Astra.03:15 Halloween sale04:08 COSMIC Desktop update05:40 New Thelio Astra arm-powered desktop07:26 Interview with Joe Speed from Ampere08:08 Joe Speed talks grassroots community09:35 Working with arm and Ampere11:48 Comparing Thelio Astra to other machines16:25 Astra as a Pre built arm desktop solution17:30 Tier IV using Ampere desktops20:58 Value in native development25:05 Running tests in parallel28:35 Britain and Tony from System76 talk about Thelio Astra's engineering29:30 Energy efficiency of Thelio Astra31:07 Britain discusses thermals32:00 Multi-threaded tasks34:50 Cooling Ampere CPUs

Support Lorenzo on Patreon.com Guest speaker: Terence McKenna PROGRAM NOTES: WARNING! If you don't want to hear Terence talking about his Timewave idea yet again, then you should skip this podcast. In this recording, Terence McKenna introduces a concept he humorously names the "Habit Reflex Increment" (HRI). He discusses the need for a unit to measure habit, reflecting on the significance of such a concept in understanding human behavior. McKenna jokes about not naming the unit after himself, contrasting his name with those of renowned scientists like Ohm or Ampere, which he finds more fitting for scientific terms. He then shifts his focus to a broader philosophical reflection, asserting that the struggle humanity faces is not eternal. He conveys an optimistic message, declaring that "novelty is winning." According to McKenna, the emergence of new ideas and innovations will ultimately prevail, bringing positive transformation and progress. Of course, you will have to listen to many more little details about the Timewave than a lot of people can put up with.

Listen as Paul Abernathy, CEO, and Founder of Electrical Code Academy, Inc., the leading electrical educator in the country, as he answers more questions from the Let's Ask Paul Portal over at www.PaulAbernathy.com.In this episode, Paul will tackle the questions submitted by a listener. The topic covered in this episode is the relationship of direct and inverse proportion to voltage, current, resistance, and power when using Ohm Law and Power Formulas. Now, I am not the BEST at electrical theory so I will try and break it down into understandable analogies to help hammer the principles home. If you are looking to learn more about the National Electrical Code, for electrical exam preparation, or to better your knowledge of the NEC then visit https://fasttraxsystem.com for all the electrical code training you will ever need by the leading electrical educator in the country with the best NEC learning program on the planet.Become a supporter of this podcast: https://www.spreaker.com/podcast/master-the-nec-podcast--1083733/support.

In these descriptions I try not to trip over hyperbolic language, but I am absolutely thrilled to have Will Killingsworth on the show today not only for his important record work at Dead Air Studios but also his musical output with bands like Orchid, Bucket Full of Teeth, Ampere and many more. He's a true lifer and someone who has focused on all of the principles that make up this DIY community and let it consume them. We are celebrating the fact that the entire Bucket Full of Teeth discography is now up on streaming services and LP courtesy of the fine folks at Iodine Recordings. Enjoy this conversation!  Listen to the Official Outbreak Podcast here (executive produced by yours truly)  Weekly Recommendation Playlist Theme Song by Tapestry Gold Subscribe to the podcast on YouTube Rockabilia sells you officially licensed Merch from ALL your favorite bands (and your Dad's favorite band, your siblings etc...). Use the promo code 100WORDSORLESS for 10% off your order.  Evil Greed is a highly curated merchandise provider from Berlin, Germany with fast, worldwide shipping and features stores from bands like Power Trip, Deafheaven, Nails, Russian Circles and so much more. Use  promo code 100WORDS for 10% off your first order today!   See omnystudio.com/listener for privacy information.