Podcasts about tmas

In this episode of Hema Now, Lucas Kühne explores the evolving landscape of immune thrombotic thrombocytopenic purpura, discussing advances in treatment, the importance of early recognition and accurate diagnosis, the growing role of real-world evidence, and why international collaboration is critical to accelerating research and improving patient care in rare thrombotic microangiopathies. Timestamps:   00:00 – Introduction  01:40 - Career inspiration and TMA interest   03:11 - Major advances in iTTP treatment   05:05 - Recognising iTTP earlier   07:08 - Distinguishing between TMAs   09:05 - Value of real-world evidence   12:13 - Future priorities in iTTP research   13:32 - Importance of research collaboration   14:57 - Advice for future researchers   15:53 - Future of TMA care 

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

Von Deutschland nach Schweden im eigenen Flugzeug: In dieser Podcast Folge sprechen wir mit Theo Voss über VFR-Flugplanung, Ostseeüberflug, schwedische TMAs, ICAO-Flugpläne und die größten Fehler, die Privatpiloten bei Flügen nach Skandinavien machen. Wir erklären, worauf man in den kontrollierten Lufträumen rund um Stockholm, Göteborg und Malmö achten muss, wie schwedische ATC arbeitet, welche Ausrüstung und Dokumente an Bord sein sollten und warum saubere Vorbereitung wichtiger ist als Improvisation im Cockpit. Außerdem geht es um Flugrouten über Dänemark und Bornholm, Transponder-Pflichten, Funkdisziplin, Moving Maps, Survival Equipment und die sichere Planung von Cross-Country-Flügen in der General Aviation.

Question of the Day:   When you open a file what are the Red Flags

I watched another budget Christmas film, and now I hate this whole idea. Basically.

A review of a gig I did that wasn't great, followed by a fourth Christmas film review that was truly terrible. Essentially.

Fourth time I've tried to record this due to stupid technical hitches. A gig review and a bit about another rubbish Christmas film on Netflix. Mostly.

Something about looksmaxxing and the 'Reviv' mouth guard (aka a myobrace) and then, a look at Hot Frosty, a very serious and important work of art. Bascially,

A kick off of some short review of Netflix Christmas films, starting with 'A Merry Little Ex-Mas' starring Alicia Silverstone. Spoilers included. It's rubbish, but they're all going to be. Basically.

Explore the upcoming TMAS 3 contract—a $916M follow-on supporting the U.S. Air Force Test Center. Learn about the scope, key incumbents, evaluation criteria, and strategies to prepare for the October 2025 RFP.Key Details:•Contract Value: $916 M•Estimated RFP Release Date: October 2025Listen to podcast to position your business for success.Contact ProposalHelper at sales@proposalhelper.com to find similar opportunities and help you build a realistic and winning pipeline.

Today's episode will focus on an update for commonly tested Thrombotic microangiopathies (TMA) such as Thrombotic Thrombocytopenic Purpura (TTP). This will include the high yield facts on mechanism, presentation, diagnostic work-up, and treatment of TTP, and also discuss Hemolytic Uremic Syndrome (HUS), and atypical HUS.

Usta gazeteciler Savaş Kerimoğlu ve Sedat Bozkurt, süreçte gelinen aşamayı, Lozan tartışmasını ve Erdoğan'ın tarihi çarpıtmasını konuşuyor. Ekonomideki gelişmeler de ikilinin gündeminde... Learn more about your ad choices. Visit megaphone.fm/adchoices

Ein Schwerpunkt liegt auf dem **italienischen Luftraumsystem**. Die italienische Struktur unterscheidet sich in Teilen deutlich von der deutschen: * Wie sind die **CTRs und TMAs** aufgebaut – insbesondere rund um **Mailand (Milano FIR)** und **Venedig (Venezia FIR)**? * Warum ihr den Funkkontakt mit FIS (Info) und ATC nicht auf die leichte Schulter nehmen solltet – und warum „Radio only“ hier nicht funktioniert. * Was es mit den **RMZ (Radio Mandatory Zones)** auf sich hat und warum ihr die **„MIL“-Zonen (militärisch)** im Auge behalten solltet. Außerdem geben wir euch praktische Tipps zu den **Flugplätzen in Italien**: * Welche Kategorien von Flugplätzen gibt es in Italien? (Aviosuperficie, Aeroporto, Campo di Volo) * Wo bekommt ihr zuverlässig **AVGAS oder Mogas** – und warum ihr am besten vorher telefonisch anfragt. * Was es mit den **PPR-Regelungen** auf sich hat und wo man alle Kontaktdaten findet.

From silly traditions to heartwarming memories, Nat, Nat & Shaun share their holiday cheer by lifting the lid on what Christmas Day looks like them. See omnystudio.com/listener for privacy information.

This webinar presents the results of the Austroads project that developed guidelines for incident response vehicles and truck-mounted attenuators. The project was a collaboration between Austroads, iMOVE, Queensland University of Technology and Deakin University. Two sets of guidelines have been developed as part of the project. The first set provides information on models/types and design specifications of incident response vehicles. The guidance focuses on three vehicle types: tow trucks, truck-mounted attenuators (TMAs) and utility vehicles. In general, the design considerations are separated into three stages: determining models and general specifications, determining additional features and equipment, and determining markings, signs and other identification measures to ensure visibility and identifiability of the incident response vehicles. The second set provides guidance on when and how TMAs and other attenuator vehicles should be used in incident response scenarios. The use of TMAs for incident response often requires departures from the established guidelines for TMA use, which are generally developed around planned operations (i.e. roadworks and maintenance activities). These issues are addressed alongside related considerations around broader temporary traffic management pertaining to incident response specifically. In the webinar, presenters Narelle Haworth, Ashim Debnath and Drew Gaynor, provide an overview of the project, including the findings of the literature review, results of the stakeholder consultations and guideline development.

Tiger, your Apecoin DAO Governance Steward | Ladies of BAYC |  Luca Netz sent us Pudgy Penguins swag | Doodles Camp Chicago | scams on OpenSea & TikTok | ZachXBT smashes TMAS | wale.swoosh's awesome market cap  timelapse | Alpha Centuri Kid | MetaMask change? | MUMBOT NFTNow's Artist of the Week | Cool Cats x Reddit | VeeFriends x Squishmallows | Fable Simulation , South Park episode generator w/AI | Links:Tiger@LadiesofBAYCMichael Keen @NFTicketJennifer Sutto @jennifer_suttoNFT Catcher Podcast @NFTCatcherPodproduced by Andy Cinquino  @ajc254NFT Catcher theme music by ItsJustLos email : NFTCatcherPod@gmail.comNFT Catcher Discord

本日のラインナップ 1. 毎週月曜日は NOA 知ってもおー漫画の日 NOA 漫画第 8 話 → https://twitter.com/akezima_d/status/1632486142212800512 あけサファ日報 → https://twitter.com/michiyoDBK/status/1632549137492131847 あけサファ部屋 → https://discord.com/channels/891212756081082389/987299197814472754 2. TMAs AL 各コレクションで配布・抽選開始 AL 配布・抽選確認 → https://discord.com/channels/895174332786049025/1059004168699912262 TMAs 前夜祭 → https://twitter.com/ninjametavelive/status/1632183759842988032 3. CNN 18 体目はイケハヤさんが 2 ETH で落札 19 体目詳細 → https://twitter.com/NinjartOfficial/status/1632299498721673216 19 体目オークション → https://cryptoninja-nouns.wtf/noun/19

Alper Ender Fırat | Muhalefetin muhalefeti dağıtması | 11.01.2023 by Tr724

It's Beginning To Look A Lot Like Sh*tmas Yes once again this show isn't available on YouTube. This time it's not technical problems, just that Ron needed a well earned rest and time to take care of his family and stuff. We are really grateful for all the messages of love and concern from you guys but please do not worry as normal service is resuming and Ron will be back better than ever. This show was kindly hosted by Ronny live in our Discord server. Very special thanks to Ronny who has gone above and beyond to help keep things running on schedule for you guys, despite having a business to run and a family to look after. Ronny we can't thank you enough you are a true legend. Also Mr. Blackman is still featured on this show on a couple of calls and I think you'll agree that a good time was had by all. You're just going to have to listen to the whole thing to see. Oh and supporter's shows are running as normal and we're doing some crazy stuff for Christmas, so there's never been a better time to sign up. And remember we guarantee laughs or your money back! Thank you as always to our kind supporters, all the people that listened live on YouTube and everyone checking out our podcast. I love you very much and keep it locked to macronshow.com where Ron will be doing more supporter's shows at BuyMeACoffee!

DiDi Richards is here to also discuss League Marketing Agreements (LMAs) in addition to her Team Marketing Agreement (TMAs) with the New York Liberty. How has she used LMAs and TMAs to open up more opportunities for herself and for the W? What does a typical off-season day look like for Richards? How did she land her own Sports Talk show on Amazon Prime? Host Jackie Powell discusses all of this, a metaphor about Trojan Horses and more with Richards. Learn more about your ad choices. Visit podcastchoices.com/adchoices

DiDi Richards is here to also discuss League Marketing Agreements (LMAs) in addition to her Team Marketing Agreement (TMAs) with the New York Liberty. How has she used LMAs and TMAs to open up more opportunities for herself and for the W? What does a typical off-season day look like for Richards? How did she land her own Sports Talk show on Amazon Prime? Host Jackie Powell discusses all of this, a metaphor about Trojan Horses and more with Richards. Learn more about your ad choices. Visit podcastchoices.com/adchoices

Giving a massive shoutout to KG in O-Town for the glorious work he did for TMAs social media presence. Talked to him about his week on TMA and what is to come from his social media stardom.

Hello Army! This past week, we were blessed with the chaotic award show Bangtan! Join us in going through the event of the TMAs, and how much laughter these men brought us! Happy listening! - The Team at Army ThinkTank. New Episode Every Thursday

“You just never know exactly who is going to know whom, so the key is to be authentic and be yourself.” ~ Pamela Slim   In this episode of Straight Talk About Sales I have Pamela Slim as my guest and we are talking Tiny Marketing Actions!  Pamela Slim is an author, speaker and business coach who works with small business owners ready to scale their businesses and IP. You may be asking what Tiny Marketing Actions are, well TMAs is a program Pamela Slim teaches that consists of small daily marketing actions, delivered consistently over a long period of time, to build your brand, business and bank account! Marketing and Sales go hand in hand like peanut butter and jelly! Join us.   Here's what you can expect from this episode:  Pamela Slim's background and the path that led her to the work she does today Leveraging Tiny Marketing Actions in our business and the benefits How to build the ecosystem specific to your business world Which TMAs will get me my desired business results How to move past the marketing roadblocks we encounter Remember to keep adding tips and strategies to your own sales toolkit that will improve your revenue outcomes and equip you to serve your clients at a higher level.    Resources mentioned in this episode:  Pamela Slim's Website The Widest Net Podcast Tiny Marketing Action course - Starts Oct 12th Preorder Selling Like A Lady    Connect with Straight Talk About Sales If you haven't done so already, subscribe to the podcast. Published episodes will come directly to your favorite podcast app.  If you enjoyed the show, please rate it on Apple Podcasts with a short review. Doing so will help me reach more entrepreneurs and small business owners just like you.  Connect with Dr. Nadia directly on LinkedIn

How about taking a little look at a classic 70s sitcom Xmas episode about life in prison with Ronnie Barker in Porridge, the merry joy of ex-footballer Chris Kamara singing some classic holiday songs, & the difficulties of Xmas during bombing raids in WWII.  Chris Kamara - Here's To Christmas

Emergencies happen in hematology and oncology. This is a fact. But how do we manage these emergencies? Look no further. In this episode, we talk all about our second hematologic emergency: disseminated intravascular coagulation (DIC) with an added bonus of an intro to thrombotic microangiopathic anemias (TMAs).Be sure to check out episode 009 on thrombocytopenia for a general approach and differential!Disseminated intravascular coagulation (DIC):Workup: CBCCMPPT, PTT, INRFibrinogenPeripheral smear - concern for schistocytes. Example of these cells from ASH image bank: https://imagebank.hematology.org/image/60306/schistocytes?type=upload#:~:text=A%20schistocyte%20is%20present%20in,angles%20and%2For%20straight%20borders.Basic mechanism of DIC is consumption of clotting factors leading to coagulopathy Need to be weary of thrombotic microangiopathy: Small blood clots forming in the small vessels leading to endothelial damage, which cause shear stress on the RBCs, which then break down into a schistocyte (AKA triangulocyte or helmet cell) Examples: thrombotic thrombocytopenic purpura (TTP) and hemolytic uremic syndrome (HUS)Management (our opinion!): - Repeat coags q4-6 hours initially (but base interval based on patient) NOTE: INR Is NOT a good assessment of “clotting status” in these situations- Repeat fibrinogen q4-6 hours initially (but base interval based on patient); keep fibrinogen >100 with cryoprecipitate in more stable patients; consider higher thresholds for more acutely ill patients (such as >150) - Repeat CBC q6-8 hours initially; can provide platelets if low, especially if they are bleeding - Workup and treatment for trigger of DIC (infection, trauma, medications, etc.)How does cirrhosis affect data interpretation?- Use clinical context to determine if labs are acutely abnormal or if they have signs/symptoms to suggest underlying liver dysfunction- In the acute setting, always just replace what is missing! How can you tell the difference between nutritional deficiencies vs. consumption (as in with DIC?)- Factor activity levels! Consider checking: Factor 8 (made in endothelium), Factor 5 (Vit K independent), Factor 7 (vitamin K dependent) - If all down, then consider DIC- If Vit K-dependent low, then nutritional deficiency Reference:https://ashpublications.org/blood/article/131/8/845/104418/How-I-treat-disseminated-intravascular-coagulation - Great How I Treat article from Blood Please visit our website (TheFellowOnCall.com) for more information Twitter: @TheFellowOnCallInstagram: @TheFellowOnCallListen in on: Apple Podcast, Spotify, and Google Podcast

TMAS is back with a vengeance! Lakes and oceans ended up being the theme in a slightly more free flowing formate. Also, our first Patreon guest Chris joins us! If you want to guest a future TMAS and share a tale of your own, consider becoming a supporter over at patreon.com/metalbreak. Enjoy!

Current exhibitionNational Hispanic Cultural CenterAlbuquerque, New MexicoThe scene:We are sitting at Cynthia's dining table inside her beautiful home, which is a sculptural display everywhere the eye travels. I'm obsessed with her kitchen, which has grey walls and ceiling to counter shelves stacked with colorful vintage ceramic Fiesta plates and bowls. The light coming from the windows is shaded and creates a sense of being enveloped inside a sacred space meant for contemplation and non-distraction. As I learn, this is intentional. Join us for our candid conversation about making a living as an artist with zero digital interaction.Highlights:+ Business name: Cynthia Cook Fine Art (no website, search her name)+ The “Terrible Mutual Admiration Society” (TMAS) TM + Cynthia has a landline and writes letters - with stamps - that's it!+ No cell phone, no texting, no email, no social media+ Talismans+ @erincurrierfineart is a super-connector and “Godess Incarnata”+ The new technology of the land and farming+ Making a living as an artist for 34 years+ Hard lessons - the best life teachers+ “I beg your pardon, I never promised you a rose garden”+ “There's a lot of tooth and claw to life”+ A description of Cynthia's art process+ Repoussé: patterns formed by hammering or pressing, esp. of metal+ Collages made of trash - “enshrining natural ephemera”+ Jewelry made from recycled metal from an origin of silver and gold-smithing+ Refusing to work with newly-mined metals+ @parsonsschoolofdesign Studio Art degree+ The expectations of the modern digital world+ “We don't want to be with anybody who doesn't want to be present with us”+ Making an effort to be with people (is not texting)+ The mark of the artist - “resisting the chicanery of the gnomes”+ Choosing a perspective of optimistic hopeful belief+ Carcass collage gifts from wild foxes+ Cynthia knew she would be a working artist at 4 years old+ The awareness and confidence of living hand-to-mouth+ The arts are a time capsule of artists sharing through the ages+ Balancing extroverted interviews and introverted nature+ Hindsight, building blocks+ “Quietly Courageous” (my biggest compliment ever!)+ Having an organic, genuine exchange with another human+ An archivist approach to documenting reality, anthropology+ Clairaudio (like clairvoyance) - when you hear sounds others can not+ Empathy and compassion and gifted sensibilities+ Caring for aging parents, with Alzheimers+ A young mind and an aging body+ Warrior-healer-goddesses-type-people+ Giving compliments is a lost art+ Doing your art with whatever medium it requires+ Live music is profound because the moment is lost as soon as you've heard it+ Most of the behaviors we have are projections of our self+ Do birds take it personally when an animal kills their young?+ The qualities of a good Boy Scout - a post-it story+ “Don't worry twice” - wait for the dataA taste:“I first encountered that saying at a job I had in college. I worked at a vintage clothing store for this fabulous goddess woman - it was kind-of a front for her drug dealing business. […] The only time she ever got in a car wreck was when she was trying to drive sober.”Favorite saying:“God grant me the serenity to accept the things I cannot change, the courage to change the things I can, and the wisdom to know the difference.” - Desiderata Support the show

Where do farts go?

Tis the season (Chris is back from harvest) and to celebrate we recorded our Filament Coffee Podcast Christmas Special!We discuss our favourite Christmas traditions and exchange gifts while sampling a Perth beer and Perth whiskey made with coffee in this festive episode.Christmas is a time for family and friends and this week we're joined by our second guest ever on the podcast, our dear friend (and family) Mr Gavin Aitkins. We've been looking forward to this one for a while.Listen in to hear what Santa brought us, to see if you can guess which of us Santa is describing in his famous"this beer is you" game and why Chris wasn't invited to Gav's wedding. That plus our favourite Christmas Day traditions and an impromptu Christmas song.Thanks so much to our listeners for supporting this new podcast this year, it's been a lot of fun and very humbling to have you tune in to our ramblings each week.Stay safe, look after each other this festive season and have a very Merry Christmas x

Mitch Lacsamana is an NFT investor, trader, and advisor. Currently, he is the Head of Marketing & Strategic Partnerships for the exclusive Discord NFT Investment Community, Metaverse HQ. Mitch also serves as an official advisor to various NFT Collections/Companies including Superlative Secret Society. ***TIMESTAMPS*** 0:00 - A long day from Miami to NYC to Jersey; Mitch's July 4th NFT Bet; The Early Days of NFTs that Mitch witnessed in 2018; Dapper Labs, NBA Top Shot & What Went Wrong 36:17 - Mitch explains Metaverse HQ & how he got involved; How Discord Works; Play & Earn Gaming; Virtual Land & Its Future Use/Value; How Metaverse HQ turned towards NFT investment/trading theme; NFT Collections and how they work; the inevitable worthless NFTs that exist in the space; CryptoPunks and why some terrible art is “worth” a lot 52:33 - “I am an NFT Degen”; Explaining floor price; The Community Aspect of NFTs, Smilesssvrs / Waheed Zai / Gio Gussen Shoutout; You never know how legit a project is; Mitch tells the story behind Mutant Cats (yes you read that right); How passive earning (staking) works with NFTs 1:08:17 - Julian asks about how some projects aren't pyramid schemes; Comparing NFTs right now to another historical market; Mitch's disciplined investment approach 1:27:30 - Discussing what notorious NFT criminal, scam artist, and insufferable scumbag, TMas, did with the recent FUD Token drop; Mitch explains the recent Wolf Game rip and what happened 1:47:20 - The hurdle of getting NFTs completely mainstream; Shaan Puri's Twitter thread on the Metaverse; Breaking down the Bored Ape Yacht Club NFT Collection and why it matters; Julian asks a pressing question on Bored Apes 2:05:47 - NFT Regulation is coming; The dangers of trading NFTs if you aren't actively paying attention; The lack of simplicity problem that continues to be one thing holding back crypto; Julian tells a story about a crypto conference in NYC back in 2018 2:24:31 - “The next great invention will look like a toy”; The silly names of NFTs and why they might actually make sense; Thinking of money in crypto and not dollars; Bitcoin (BTC), Ethereum (ETH), and the long term look of the crypto space; Mitch talks about what the DeFi guys were telling him at Art Basel in Miami; Crypto, DeFi, and the entire space is still so new 2:42:55 - Web3: What is it and what will it look like; Mitch explains why he's still unsure; The Superlative Secret Society NFT Collection that Mitch advises for; Axie Infinity in the Philippines; Julian discusses Mitch's brother Mike and Don Tapscott Ted Talk he showed him that started everything ~ YouTube EPISODES & CLIPS: https://www.youtube.com/channel/UC0A-v_DL-h76F75xik8h03Q  ~ PRIVADO VPN FOR $4.99/Month: https://privadovpn.com/trendifier/#a_aid=Julian   Get $100 Off The Eight Sleep Pod Pro Mattress / Mattress Cover: https://eight-sleep.ioym.net/trendifier  Julian's Instagram: https://www.instagram.com/julianddorey  ~ Beat provided by: https://freebeats.io  Music Produced by White Hot

In this weeks' recap on Twitter Spaces we dive into: Tmas fud token Big Doodle sales ConstitutionDAO Defi Kingdoms Metaverse wars Pirate bay nfts Listen in and stay tuned for Next weeks' recap every Thursday night at 9 PM EST DISCLAIMER: All of the information discussed in our podcast is for entertainment purposes only. As with any financial endeavor, do your own research. --- This episode is sponsored by · Anchor: The easiest way to make a podcast. https://anchor.fm/app --- Send in a voice message: https://anchor.fm/non-refungible/message Support this podcast: https://anchor.fm/non-refungible/support

Paradox House presents… Episode 7 of Scripted hosted by Daisy Lewis. This week Daisy sat down with one of the U.K's brightest writers in Anya Reiss to talk through all things scripted. We also bump up the excitement with her upcoming TV show, ‘Becoming Elizabeth', that is being distributed via STARZ! Anya Reiss began her writing career in theatre with her debut play Spur of the Moment at the Royal Court Theatre in 2010. She won the Most Promising Playwright Award at both the Critics Circle and Evening Standard awards that year along with Best New Play at the TMAs. Her follow up play The Acid Test was staged at the same venue the next year and her National Theatre Connections play Forty-Five Minutes was in 2013. Her original version of The Seagull, directed by Russell Bolam, was staged in 2012 at Southwark Playhouse, and they worked on two further modern-day Chekhovs together at the same venue and then St James Theatre. Since then her version of Spring Awakening toured with Headlong and an adaptation of Oliver Twist was at the Regents' Park Theatre in 2017. Anya has worked in television, a core writer on Eastenders and a lead writer on series one of Channel 4's Ackley Bridge. She is currently writer-producer on Starz's Becoming Elizabeth which will air next year. Enjoy!

Hey ARMYs! This week we go over BTS at the TMAs, our excitement about the Permission to Dance on Stage offline concerts, and TxT's first online concert ACT:BOY!Make sure to let us know your thoughts! Happy listening! - The Team at Army ThinkTank. New Episode Every Thursday

After deep diving into David Whyte and spending more time learning about astrology, Hillary adds a whole new layer to her definition of identity and the role it plays with storytelling. Listen to Hillary's conversation with astrologer and poet Heidi Rose Robbins and explore the idea of asking more beautiful questions of yourself and others and how storytelling can do just that!Additional reading/listening:Click here to learn more about Heidi's astrology offerings, including the Progressed Moon workbook.Listen to the episode of Heidi's podcast, The Radiance Project, where she dives into the Progressed Moon and her own stories around it. Listen to Heidi's interview with Hillary on The Radiance Project.Listen to Heidi's interview with her dad on The Radiance Project.Listen to Heidi's interview with her mom on The Radiance Project.Follow Heidi on InstagramDavid Whyte's book Consolations Dear Listener,Make sure you listen to all of the previous Season 3 episodes for listening to this one. This is a serialized story. Thanks,Hillary Rashomon is produced and hosted by Hillary ReaAdditional music in this episode is by Liz DeliseLiz Delise's websiteLiz's band LizdeliseWatch the music video for NobodyRashomon theme music is by Ryan Culinane courtesy of the Free Music ArchiveRashomon's album art is by Thom LessnerThis episode of Rashomon is sponsored by Tell Me A Story, a communication consulting and coaching business that trains entrepreneurs, leaders and change makers to use the art of storytelling as a powerful communication tool. Learn more about working with Hillary 1-on-1 in one of TMAS' Crafting Your Narrative programs.

SPOILER ALERT: Please watch In & Of Itself on Hulu before listening to this episode. After watching Derek Delgaudio's In & Of Itself on Hulu, Hillary begins to question whether or not storytelling an illusion of identity. Plus she regrets not having the foresight to see In & Of Itself when it was Off-Broadway in New York City. In hopes of making up for the lost experience, Hillary interviews Priya Ollapally Wellington, an audience member from when the show was Off-Broadway at the Daryl Roth theater. Priya shares her experience of selecting her I AM card and they talk about magic, identity, and how storytelling plays a part. A huge thanks to Priya Ollapally Wellington for having this conversation. And my apologies to Derek Delgaudio for mispronouncing your last name throughout the episode! Dear Listener,Make sure you listen to all of the previous Season 3 episodes for listening to this one. This is a serialized story. Thanks,Hillary Rashomon is produced and hosted by Hillary ReaAdditional music in this episode is by Liz DeliseLiz Delise's websiteLiz's band LizdeliseWatch the music video for NobodyRashomon theme music is by Ryan Culinane courtesy of the Free Music ArchiveRashomon's album art is by Thom LessnerThis episode of Rashomon is sponsored by Tell Me A Story, a communication consulting and coaching business that trains entrepreneurs, leaders and change makers to use the art of storytelling as a powerful communication tool. Learn more about working with Hillary 1-on-1 in one of TMAS' Crafting Your Narrative programs. Thank you for listening! If you enjoyed this episode, please share with a friend. You can listen to all of our past seasons from our Simplecast page. 

Hillary takes every personality test on the market and digs in to Clifton Strengths with Surabhi Lal.At the beginning of the pandemic, Hillary took many personality tests to see if it changed anything about her identity.  Her findings were increased frustration, doubt and skepticism. In January of 2021 Hillary was forced to take yet another personality test called Clifton Strengths. This led to even more questions so she reached out to Clifton Strengths certified coach and business consultant, Surabhi Lal. A huge thanks to Surabhi Lal for having this conversation. Learn more about Surabhi's work on her website: https://www.surabhilal.com Dear Listener,Make sure you listen to all of the previous Season 3 episodes for listening to this one. This is a serialized story. Thanks,Hillary Rashomon is produced and hosted by Hillary ReaAdditional music in this episode is by Liz DeliseLiz Delise's websiteLiz's band LizdeliseWatch the music video for NobodyRashomon theme music is by Ryan Culinane courtesy of the Free Music ArchiveRashomon's album art is by Thom LessnerThis episode of Rashomon is sponsored by Tell Me A Story, a communication consulting and coaching business that trains entrepreneurs, leaders and change makers to use the art of storytelling as a powerful communication tool. Learn more about working with Hillary 1-on-1 in one of TMAS' Crafting Your Narrative programs. Thank you for listening! If you enjoyed this episode, please share with a friend. You can listen to all of our past seasons from our Simplecast page. 

What's the difference between personality and identity? Hillary meets with Dr. Jasara Hogan to explore this question and learn more about her relationship with Millicent Wimbleberry, epigenetics and the problem with personality. Dear Listener,Make sure you listen to all of the previous Season 3 episodes for listening to this one. This is a serialized story. Thanks,Hillary Rashomon is produced and hosted by Hillary ReaAdditional music in this episode is by Liz DeliseLiz Delise's websiteLiz's band LizdeliseWatch the music video for NobodyRashomon theme music is by Ryan Culinane courtesy of the Free Music ArchiveRashomon's album art is by Thom LessnerThis episode of Rashomon is sponsored by Tell Me A Story, a communication consulting and coaching business that trains entrepreneurs, leaders and change makers to use the art of storytelling as a powerful communication tool. Learn more about working with Hillary 1-on-1 in one of TMAS' Crafting Your Narrative programs. Thank you for listening! If you enjoyed this episode, please share with a friend. You can listen to all of our past seasons from our Simplecast page. 

Hillary and Millicent share a love of Disneyland and dream of a future where it's possible to live there. Hillary reads the final chapter of Millicent Wimbleberry: The Early Years and imagines what it would be like in the after life if her family could come back together and ride a log flume in the happiest place on earth (and beyond). Dear Listener,Make sure you listen to all of the previous Season 3 episodes for listening to this one. This is a serialized story. Thanks,Hillary Rashomon is produced and hosted by Hillary ReaAdditional music in this episode is by Liz DeliseLiz Delise's websiteLiz's band LizdeliseWatch the music video for NobodyRashomon theme music is by Ryan Culinane courtesy of the Free Music ArchiveRashomon's album art is by Thom LessnerThis episode of Rashomon is sponsored by Tell Me A Story, a communication consulting and coaching business that trains entrepreneurs, leaders and change makers to use the art of storytelling as a powerful communication tool. Learn more about working with Hillary 1-on-1 in one of TMAS' Crafting Your Narrative programs. Thank you for listening! If you enjoyed this episode, please share with a friend. You can listen to all of our past seasons from our Simplecast page. 

Hillary explores how being an only child and being alone are two completely different ways of being. You can't choose to be an only child but you can choose whether or not you are alone and whether or not there are family stories worth telling. Plus, Chapter Five of Millicent Wimbleberry: The Early Years is a dramatic, fictional story about Hillary's real life cousin. (All of the names have been changed to protect the innocent.) And yes, there is another David Whyte quote to enjoy. Dear Listener,Make sure you listen to all of the previous Season 3 episodes for listening to this one. This is a serialized story. Thanks,Hillary Rashomon is produced and hosted by Hillary ReaAdditional music in this episode is by Liz DeliseLiz Delise's websiteLiz's band LizdeliseWatch the music video for NobodyRashomon theme music is by Ryan Culinane courtesy of the Free Music ArchiveRashomon's album art is by Thom LessnerThis episode of Rashomon is sponsored by Tell Me A Story, a communication consulting and coaching business that trains entrepreneurs, leaders and change makers to use the art of storytelling as a powerful communication tool. Learn more about working with Hillary 1-on-1 in one of TMAS' Crafting Your Narrative programs. Thank you for listening! If you enjoyed this episode, please share with a friend. You can listen to all of our past seasons from our Simplecast page. 

Chapter Four of Millicent Wimbleberry: The Early Years takes a dark turn with the sudden death of a pet bird and a French teacher.  Murder aside, this story feels the most like a true Hillary story than all of the previous chapters of her childhood memoir. Hillary explores school stories related to science experiments involving pancakes, lizards, and urine. It is in these science stories that Hillary starts to discover the stories that she wants to share with a live audience.  Dear Listener,Make sure you listen to all of the previous Season 3 episodes for listening to this one. This is a serialized story. Thanks,Hillary Rashomon is produced and hosted by Hillary ReaAdditional music in this episode is by Liz DeliseLiz Delise's websiteLiz's band LizdeliseWatch the music video for NobodyRashomon theme music is by Ryan Culinane courtesy of the Free Music ArchiveRashomon's album art is by Thom LessnerThis episode of Rashomon is sponsored by Tell Me A Story, a communication consulting and coaching business that trains entrepreneurs, leaders and change makers to use the art of storytelling as a powerful communication tool. Learn more about working with Hillary 1-on-1 in one of TMAS' Crafting Your Narrative programs. Thank you for listening! If you enjoyed this episode, please share with a friend. You can listen to all of our past seasons from our Simplecast page. 

With a little bit of help from Philip Pullman, Millicent Wimbleberry Chapter Three, and her more recent pet stories, Hillary explores the question: Do we project our inner-self onto our animals? Dear Listener,Make sure you listen to all of the previous Season 3 episodes for listening to this one. This is a serialized story. Thanks,Hillary Rashomon is produced and hosted by Hillary ReaAdditional music in this episode is by Liz DeliseLiz Delise's websiteLiz's band LizdeliseWatch the music video for NobodyRashomon theme music is by Ryan Culinane courtesy of the Free Music ArchiveRashomon's album art is by Thom LessnerThis episode of Rashomon is sponsored by Tell Me A Story, a communication consulting and coaching business that trains entrepreneurs, leaders and change makers to use the art of storytelling as a powerful communication tool. Learn more about working with Hillary 1-on-1 in one of TMAS' Crafting Your Narrative programs. Thank you for listening! If you enjoyed this episode, please share with a friend. You can listen to all of our past seasons from our Simplecast page. 

As long as she can remember, Hillary has always declared (to anyone who will listen) that she hates tomatoes. But that's not the full story. In this episode Hillary explores old photographs, looking for the true stories contained within. You'll also hear Millicent Wimbleberry: Chapter Two and the perspectives of Hillary's parents, Connie and Steven. If you haven't yet listened to the first two episodes of this season of Rashomon, you need to listen to them before listening to this one. Otherwise you will miss most of what is happening with Hillary's childhood memoir Millicent Wimbleberry: The Early Years and the audio assignment she gave to her parents. Additional reading/listening:The photo from PunkHouse Philly's InstagramAtom and His Package Rashomon is produced and hosted by Hillary ReaThank you to Adam Goren aka Atom and His Package for giving us permission to play his song "Threshold to Adulthood" in this episode. (Sorry for calling it 'Thresholds'.)Additional music in this episode is by Liz DeliseLiz Delise's websiteLiz's band LizdeliseWatch the music video for NobodyRashomon theme music is by Ryan Culinane courtesy of the Free Music Archive This episode of Rashomon is sponsored by Tell Me A Story, a communication consulting and coaching business that trains entrepreneurs, leaders and change makers to use the art of storytelling as a powerful communication tool. Learn more about working with Hillary 1-on-1 in one of TMAS' Crafting Your Narrative programs. Thank you for listening! If you enjoyed this episode, please share with a friend. You can listen to all of our past seasons from our Simplecast page. 

Hillary confesses that she had a codependent relationship with Jessica Simpson's autobiography Open Book towards the beginning of quarantine. And thanks to Jessica, Hillary was inspired to turn her own childhood memoir into an audio book. Millicent Wimbleberry: The Early Years begins with "The Avocado Accident" a story that pre-dates Hillary's memory. Hillary searches for deeper meaning to this story and wonders if baby stories are not her stories to tell, rather they belong to her parents. Speaking of... you'll hear from them as well. Friends - make sure to listen to the Prologue and Episode 1 of this season to have a better understanding of the full story. Additional reading/listening:Listen to Open Book by Jessica SimpsonListen to these episodes of the podcast You're Wrong About Quarantine Deep Dive: Jessica Simpson’s “Open Book” (Week 1)Quarantine Deep Dive: Jessica Simpson’s “Open Book” (Week 2)Quarantine Deep Dive: Jessica Simpson’s “Open Book” (Week 3)Quarantine Deep Dive: Jessica Simpson’s “Open Book” (The Conclusion!)Read Consolations by David WhyteListen to this episode of On Being with Krista TippettThe Conversational Nature of Reality an interview with David Whyte Rashomon is produced and hosted by Hillary ReaMusic in this episode is by Liz DeliseLiz Delise's websiteLiz's band LizdeliseWatch the music video for NobodyTheme music is by Ryan Culinane courtesy of the Free Music Archive This episode of Rashomon is sponsored by Tell Me A Story, a communication consulting and coaching business that trains entrepreneurs, leaders and change makers to use the art of storytelling as a powerful communication tool. Learn more about working with Hillary 1-on-1 in one of TMAS' Crafting Your Narrative programs. Thank you for listening! If you enjoyed this episode, please share with a friend. You can listen to all of our past seasons from our Simplecast page. 

Hillary discovers her old My Book About Me and begins to question the book's significance. Did she really march to the beat of her own drum as a kid, like she remembers? Or did she choose the status quo by filling out the pages of this store bought book like everybody else? My Book About Me is a book written by Dr. Seuss and RoyMckee.Learn more about Dr. Suess' Anti-Black and Anti-Asian Published Work:Read this post from The Conscious Kid Instagram: @theconsciouskidRead Dr. Seuss Books Are Pulled, and a ‘Cancel Culture’ Controversy Erupts (The NY Times)Read "Know the difference between canceled and accountability" (Anti-Racism Daily) Rashomon is produced and hosted by Hillary ReaMusic in this episode is by Liz DeliseLiz Delise's websiteLiz's band LizdeliseWatch the music video for NobodyTheme music is by Ryan Culinane courtesy of the Free Music Archive This episode of Rashomon is sponsored by Tell Me A Story, a communication consulting and coaching business that trains entrepreneurs, leaders and change makers to use the art of storytelling as a powerful communication tool. Learn more about working with Hillary 1-on-1 in one of TMAS' Crafting Your Narrative programs. Thank you for listening! If you enjoyed this episode, please share with a friend. You can listen to all of our past seasons from our Simplecast page. 

Heute reden wir mit euch über die End Of The Year Award Shows, zu denen unter anderem die sehr wichtigen Melon Music Awards und die Mnet Asian Music Awards gehören.

''Neden Müslümanım?'' kitabımın 1. bölümü.

Peter and Chris virtually perform their hit fringe festival show “A Peter N' Chris-tmas Carol”.  Enjoy this during the Holidays for a limited time!  One day when we can perform live again this will be taken down, so we can get your money. UNTIL THEN ENJOY AND HAPPY HOLIDAYS!! Original music by Devon Hyland! FOLLOW US ON INSTAGRAM: @ttidpod/ @PeterNChris / @imchriswilson / @Petercarlone FOLLOW US ON TWITTER: @PeterNChrisShow / @imchriswilson / @petercarlone Brought to you By: The Sonar Network

it's christmas eve but this episode has basically nothing to do with that! the ladies discuss cheesy chicken, a cabbage sal, and music, because we make the rules around here!!

Christmas is about spending time with family & friends - it's one of the most joyous ways to celebrate! As proof, Jason & Chrissy are joined by Adam Simmons and Chris Wutzke for an embarrassing discussion filled with tragic language and unfortunate jokes.What are your fondest Christmas gifts?Which movies do you watch every year?What are the best ways to connect with your loved ones?We're here to sh*t on all that, and anything else you cherish.Check out Liquid Death Mountain Water & use code ‘PARTYNAKED’ at checkout to save 10%!More hilarious podcasts await you at The Inner Circle Podcast Network so go check em out!Tweet us!Email us! jason.alme@teamalme.com

As promised, the Babblement returns for it's Christmas Special. As you can imagine, joy is all around us. On the Babble Christmas list this year is; Nigella's Microwave, Robert Dyas and Keith Chegwin on the Wii. We really want to hear from you! If you'd like to contribute a Hate List or to be featured on future episodes of the Babblement, you can get in contact in all the following ways; Instagram and Twitter = @babblementpodFacebook = fb.me/babblmentpod Email = babblementpod@gmail.com If you're feeling extra nice, rate and subscribe on iTunes - you could be in with the chance of winning something a Babblement T Shirt!

Link to bioRxiv paper: http://biorxiv.org/cgi/content/short/2020.11.19.390401v1?rss=1 Authors: Geread, R. S., Sivanandarajah, A., Brouwer, E., Wood, G., Androutsos, D., Faragalla, H., Khademi, A. Abstract: In this work, a novel proliferation index (PI) calculator for Ki67 images called piNET is proposed. It is successfully tested on four datasets, from three scanners comprised of patches, tissue microarrays (TMAs) and wholeslide images (WSI), representing a diverse multicentre dataset for evaluating Ki67 quantification. Compared to state of the art methods, piNET consistently performs the best over all datasets with an average PI difference of 5.603%, PI accuracy rate of 86% and correlation coefficient R = 0.927. The success of the system can be attributed to a number of innovations. Firstly, this tool is built based on deep learning, which can adapt to wide variability of medical images and it was posed as a detection problem to mimic pathologists workflow which improves accuracy and efficiency. Secondly, the system is trained purely on tumour cells, which reduces false positives from non-tumour cells without needing the usual pre-requisite tumour segmentation step for Ki67 quantification. Thirdly, the concept of learning background regions through weak supervision is introduced, by providing the system with ideal and non-ideal (artifact) patches that further reduces false positives. Lastly, a novel hotspot analysis is proposed to allow automated methods to score patches from WSI that contain significant activity. Copy rights belong to original authors. Visit the link for more info

FREE-TOBER is in full effect. We covered a lot of ground in this episode. Brand new TMAS from Ace, A 2 week champion of our Battle of Bands, a new challenger and Odd reads a selection from J. W. Ocker's new book Cursed Objects. The book is available on Amazon.All that and so much more.O.T.I.S. FBOlio (BoB 2 week champ)Liliac (Challenger)

Another great addition to our Free-Tober offerings. With Odd this time is an old friend and previous co-host Mr. Dr. Cookie. Cookie and Odd have been friends for a while and always have a good laugh or 12 together.We have a new #TMAS from a listener as well as a new champ in our battle of bands contest. Our new winner is Liliac, a young Rock band made up of 3 brothers and 2 sisters. They do covers and original songs, they blew the competition out of the water and due to Odd taking next week off. They'll enjoy a 2 week reign as champ...unchallenged.At the top of the show, Odd talks about a friend of his who is in need of help. Odd's friend Liz has a daughter who is very sick and the doctors around them are baffled as to what is wrong. Audrey has been sick off and on her whole life, it's some type of auto immune disorder but nobody knows exactly what. Most recently she had a devastating flare-up and has been down for several days. Worse than anything they had dealt with previously, Liz has reached out to the Mayo Clinic and they are willing to see, diagnose and treat Audrey...unfortunately despite Liz's husband having great insurance....they refuse to cover any of the impending costs. They have set up a Go Fund Me to try to get help in covering the 14,000 needed. Please, donate or share...anything helps.Audrey's Go Fund Me:https://gf.me/u/y4vtrtLiliac FB:https://www.facebook.com/liliacbandLiliac Website: https://www.liliacband.com/liliacband

Pam Slim, owner of Main Street Learning Lab, once again joins Small Biz Buzz to discuss how little marketing actions make it easier to take action in your small business. Slim recommends that small business owners leverage ecosystem development and partner development to look at ways in which they can understand their bigger systems. “I call [it] tiny marketing actions, which can definitely be marketing actions where you're planting the seed to get the word out, where you get more visibility, where you connect with customers, but it's also really tiny relationship actions,” said Slim. “There are ways that you're just slowly in little tiny ways connecting with people and connecting people with each other, and it's when you do that, at first it can take a little bit of time to learn the different players, but over time as you start to do that weaving, it begins to generate a certain kind of momentum.” Pam suggests that small business owners ask, "What are tiny little ways that I can start to plant the seed, connect with somebody, reach out, ask a question, make a connection?" and then make it a habit. That's the real key–make that the way you operate on a daily basis, and that's where you’ll find a lot of momentum in your marketing. Click play for more.

What if you wanted to download a movie to watch on your next flight, and instead of taking several minutes on the airport WiFi, it could take seconds on your mobile device? 5G technology is set to make that happen, but first the infrastructure to support this next generation of network connectivity needs to be deployed.That's where FDH comes in, putting its expertise to work by helping service providers deploy and maintain their infrastructure. So eventually you won't have to wait to download that movie.“You see antennas and radios frequently being swapped out,” says Dana Clauson, AE Services Manager at FDH. “Most of the time, they're doing away with TMAs, which are tower-mounted antennas, and radios are being decommissioned, and they're replacing them with integrated antennas, antennas that have radios built into them, that have the ability to use multiple different bands the carriers need.”“These are many of the things we're seeing over time on the macro side,” he continues. “And … a lot of small cells are also being added in to supplement those bigger, more densely populated areas that really require the bandwidth the networks are trying to support.”The transition to 5G requires not only improvement to technology, but also an understanding of the concerns and needs of local communities and the government officials who represent them during the rollout process. “One of the changes with 5G is our clients, the carriers and the tower owners – they're very used to just replacing equipment that's on the tower. With this new 5G technology, there are these additional steps of running the fiber and finding different locations for the equipment to go on that's not a standard tower,” says Krystyn Perez, PE, FDH Vice President of Engineering. “There are a lot of jurisdictions that were not originally a big fan of having a new panel on every light pole in the city every 200 feet, so a lot of those things have come up with the 5G change.”Perez adds, “We've taken a lot of steps to make sure we understand the technology, understand what the jurisdictions are going to require for approval of these installations, and understand what the carriers are trying to do with their site locations so we can do what we need to help them get their site installations approved.”“We get calls all the time with challenges that seem insurmountable to our clients,” says Chad Barham, Director of Client Services. “A lot of our efforts on the client services side are defining what the clients' biggest issues are and defining ways to overcome them.”

Kickbackwithchris - tmas by Chris Jones

Get cozy and celebrate the holidays with the SLT crew! We'll swap traditions, learn about the Krampus, exchange gifts and more! Citations : https://docs.google.com/document/d/1pONVwKh1WIxfQoROfUFGQlnKb8lE33Rc7UU8Va26e5Y/edit?usp=sharing (https://docs.google.com/document/d/1pONVwKh1WIxfQoROfUFGQlnKb8lE33Rc7UU8Va26e5Y/edit?usp=sharing)

In this very special episode of Emergency Exit, producer Brandon is sitting in for both Carlos and Jimmy as the are headed home for Christmas, or is it for the holidays? That's what Brandon wants to figure out, is the phrase Merry Christmas decreasing in use and a few other Christmas topics. 

The dumbasses are back baby!!! Yeah um sorry about dropping off the planet for, like, a whole year without saying anything but in our defense... we have absolutely no excuses. So to catch you up, Hope got married and Brandi went on a trip around the world! And now we’re back with a spooky Christmas sequel, introducing merry horse skulls and biblical yule witches. Yes, Mari Lwyd and La Befana are here and you better let them into your house. Now. And if you want to see what Brandi was up to during the break, check out Art Sistory! It’s her new comedy Art History podcast about dishing goss and trying to be alive! https://podcasts.apple.com/us/podcast/art-sistory/id1459472985 and visit https://www.instagram.com/artsistory/

Alexxus and Laken are joined by their producer Caleb on this episode of MOO, examining the past year and looking forward to what’s ahead.

Whats Brewin? Peleton Commercial Mcdonalds Chicken Sandwich What You Missed: Jay-Z Barron Trump Uni Report Remember that one time: Chris is now on TikTok Last Call: Our wish list

One at One | John Conway LIVE! | The Chair Challenge | Deep Thought of the Week | 4 Days of Shitmas | Olly's Phone Fail | Tom's Been Everywhere Instagram | Facebook | Twitter | Hit App

Dr. Lewis discusses TMAs in the post-BMT setting, recent research into the topic, when to think about it and what to do about it.

In this podcast military spouse Lynda MacFarland and her daughter Maggie Phillips share insight on open communication in their family and share some uplifting stories about the military life experience.   Register for the MCEC National Training Seminar (NTS) being held July 23-25 at Renaissance Washington DC Downtown Hotel by visiting our website at: www.militarychild.org To learn more about Drowning in Lemonade check out Lynda’s blog at:  https://drowning-in-lemonade.com/   School Liaison Officers (SLOs) are the primary point of contacts on military installation for school-related issues, and exist to assist families whose children’s education is affected by military life. https://branchta.org/role-school-liaison-officer-slo/   Student to Student Program MCEC provides support to military-connected and civilian students through our student-led, peer-to-peer mentoring programs at the K-12 levels. Elementary Student 2 Student™ (eS2S™), Junior Student 2 Student® (JS2S™), and Student 2 Student® (S2S™) serve the purpose to ease transitions, and create a positive environment for any new student. To hear more about this program visit our website and click on the audio link: https://www.militarychild.org/audience/students   Tell Me a Story The MCEC Tell Me A Story® program is an initiative created to empower our military connected children by using literature and their own stories. Stories have the capacity to open family discussions on potentially difficult topics, such as family separation, deployment, moving, grief, and crisis. To find out more or to initiate a TMAS event in your community visit: https://www.militarychild.org/programs/tell-me-a-story-tmas  

I didn’t even know we had it lmao! --- Support this podcast: https://anchor.fm/nickdrivas/support

Welcome to Tell Me A Story, the newest show to come out of UNC Asheville's student radio station, Blue Echo. We'll be airing new stories from real life and from fiction and exploring how we communicate our experiences, hopes, and dreams. TMAS airs every Wednesday from 4-5pm on http://echo.blueecho.unca.edu/, but we'll also be publishing the show here on our podcast feed so you can listen to it whenever! What's your story?

In the latest episode of the Win In 6 Podcast, Behind The Buck Pass Site Expert Adam McGee is joined by contributor Jordan Treske to discuss Khris Middleton's recovery timeline, the Cleveland Cavaliers' comments and much more in a special Christmas Day podcast.

Rick has returned to tell of his adventures. Mariya and Spencer had adventures of their own in Portland, Oregon during Bridgetown Comedy Festival. This is a classic TMAS episode for the whole dysfunctional family. 

http://facebook.com/15thSineOfficial | http://facebook.com/revampedproductions | http://faceboook.com/revampedglobaltrance | http://www.revampedproductions.com.au | Every Sunday Night on TFB-Radio 10:00 PM - 12:00 Midnight (Melbourne, Australian Time). Listen link- http://bitly.com/tfb_radio 01. 15th SINE - The Beginning Of A Trance Journey (Intro 001) 02. Alexander Popov - Multiverse 03. Protoculture - Pegasus 04. Marc Simz - Submarine 05. Mark Sherry - The Pillars Of Creation 06. Kinetica - The Last Wish 07. Tangle - Sahara (A.R.D.I. Remix) 08. UCast & George Kamelon - Jump 09. Alexander Turok - Oscillation (Dart Rayne & Yura Moonlight Remix) 10. Digital X - Legioner 11. Corti Organ - Butterfly 12. Jorn van Deynhoven - Freaks (Festival Mix) 13. Signum - What Ya Got 4 Me (Duncan Newell Remix) 14. Artic Moon - Into The Dusk 15. Cosmic Gate & JES - Yai (Here We Go Again) 16. Armin van Buuren presents Rising Star feat Betsie Larkin - Safe Inside You 17. Ost & Meyer, Ronski Speed And Cate Kanell - Fortress (Dan Stone Remix) 18. Talla 2XLC & RAM feat. Kim Kiona - Untill The End 19. Armin van Buuren - Save My Night (Andrew Rayel Remix) 20. Tigerlily & The Only - Daylight - (15th SINE Melodic Trance NRG Mix) 21. DJ Tim & Misjah - Access (John Askew Remix) 22. Svenson & Gielen - Twisted (Jorn van Deynhoven Remix) 23. Orjan Nilsen - Shananigans 24. Mark Sixma - Vendetta 25. 15th SINE - Altered States Of Consciouness

Mariya wants to meet Stephen Hawking to discuss the secrets of the universe and fucking. A hilarious clip from The Bachelor inspires another great piece of TMAS improv. The gang watches a demonstration on a sexual application of grapefruit. A fan named his fish after Mariya. And more!

Hey #sydneytrancefam, it's time for this week's instalment of Sydney Trance. I will be at the helm this week with a recreation of my 5am closing set @ Progadelic. Big shouts out to everyone that made it down to support our amazing local talent and a special mention to those of you that rocked out with me until close. #trancefamily #psydney Tracklist: 1) Son Kite - Let Us Be (Loud Remix) 2) Sonic Species - Zero (Outsiders Remix) 3) Deedrah vs Talamasca - Last Debate 4) Vertical Mode - Madness Express 5) Mystical Complex - Now is the time 6) Splattered Implant & Zyrus 7 - Brainstorm (Splattered Implant darker mix) 7) Phanatic - Psychedlic Science 8) Astrix - Type 1 (Harmonic Rush Remix) 9) Ikerya Project - Morning Exercise (Autocinema Remix) 10) Ital, 2012 - Irmaos 11) The Big Bang - Blaster 12) A-Team - Hadrush Have you joined our FaceBook Group? Sydney Trance Scene Have you joined our FaceBook Page? Sydney Trance Find us on iTunes! Sydney Trance Itunes Please take a moment to check out our sponsors! Critical Overload Komplex Sounds TMAS TrancenDence

Hey #sydneytrancefam, it's time for this week's instalment of Sydney Trance. Up this week we have a special live guest mix, recorded live at ST pres Coming Soon on the 27.09.2014 @ Miind Nightclub, from Galaktik, Enjoy! Tracklist: Sorry no tracklist. Have you joined our FaceBook Group? Sydney Trance Scene Have you joined our FaceBook Page? Sydney Trance Podcast Find us on iTunes! Sydney Trance Itunes Please take a moment to check out our sponsors! Sydney Underground Latest Trance TMAS TrancenDence

Hey #sydneytrancefam, it's time for this week's instalment of Sydney Trance. Up this week we have a special live guest mix, recorded live at ST pres Coming Soon on the 27.09.2014 @ Miind Nightclub, from Mr ST, Zac Slade, Enjoy! Tracklist: Sorry no tracklist. Have you joined our FaceBook Group? Sydney Trance Scene Have you joined our FaceBook Page? Sydney Trance Podcast Find us on iTunes! Sydney Trance Itunes Please take a moment to check out our sponsors! Sydney Underground Latest Trance TMAS TrancenDence

Hey #sydneytrancefam, it's time for this week's instalment of Sydney Trance. Up this week we have a special live guest mix, recorded live at ST pres Coming Soon on the 27.09.2014 @ Miind Nightclub, from Pato De Gomah, Enjoy! Tracklist: Sorry no tracklist. Have you joined our FaceBook Group? Sydney Trance Scene Have you joined our FaceBook Page? Sydney Trance Podcast Find us on iTunes! Sydney Trance Itunes Please take a moment to check out our sponsors! Sydney Underground Latest Trance TMAS TrancenDence

Hey #sydneytrancefam, it's time for this week's instalment of Sydney Trance. Up this week we have a guestmix from Mr Versatile, TONTO. Tonto has put together a 1 hour guest mix for your listening pleasure. Enjoy! Tracklist: 1. Phaxe - Street lights (Original Mix) 2. Ticon - Hops of Hades (Original Mix) 3. Coming Soon - Become One (Interactive Noise Remix) 4. Freedom Fighters - Headroom (Tribelistic Lifeforms Remix) 5. Major 7 & D-Addiction - Drugs (The Remix) 6. Mark Sherry & 3DW vs Madders feat Debbie Sharp - Feel So Right (Original Mix) 7. D-Addiction - WTF (Coming Soon Remix) 8. Major 7 & Vini Vici - Back Underground (Original Mix) 9. Morten Granau & Vice - The Pressure (Original Mix) 10. Major 7 ft. Black & White - Black 7 (Coming Soon Remix) 11. Jordan Suckley & Paul Webster - HELP! (Original Mix) 12. Dave Spoon - At Night (Phillip Estevez Remix) Have you joined our FaceBook Group? Sydney Trance Scene Have you joined our FaceBook Page? Sydney Trance Podcast Find us on iTunes! Sydney Trance Itunes Please take a moment to check out our sponsors! Sydney Underground Latest Trance TMAS TrancenDence

Hey #sydneytrancefam, it's time for this week's instalment of Sydney Trance. Up this week we have a guestmix from our resident DJ, VJ & Graphic Designer Pato De Gomah. PDG has put together a 1 hour 20 guest mix for your listening pleasure. Enjoy! Tracklist: 1. Spencer Brown - Chalice (Original mix) 2. Evave - Falling in My Eyes (Original Mix) 3. Kris O'Neil, Max Freegrant - The Dark Passenger (Yuji Ono Remix) 4. Drumcomplex & Roel Salemink - The Box (Alex Di Stefano Remix) 5. Alex Di Stefano - Phoneutria (Album Edit) 6. Matt Minimal - Krank 13 (Skober Remix) 7. Matt Lange - Bad Year Blimp 8. Airwave - The Wrath Of Tambora (Simon Templar Remix) 9. Alex Di Stefano - Back Again 10. The Digital Blonde - Noc2One (Midnight Mix) 11. Royal Flush – Definition of Insanity 12. Ace Ventura - Presence (Interactive Noise Remix) 13. Talpa - Going Home 14. Klopfgeister and Capital Monkey - Tomorrow Have you joined our FaceBook Group? Sydney Trance Scene Have you joined our FaceBook Page? Sydney Trance Podcast Find us on iTunes! Sydney Trance Itunes Please take a moment to check out our sponsors! Sydney Underground Latest Trance TMAS TrancenDence

Hey Sydney Trance Family, It's time for this week's instalment of Sydney Trance. Up this week we have a Steve Ari. This is an absolute cracker of a mix so sit back relax and enjoy. Track list: valley of the kings - claudia cazacu the void - ReOrder, darren porter physical overdrive - johan gielen high on mel - astrix mad - coming soon, vini vici unleash - sean tyas thesound of goodbye at the airport - photographer vs armin van buuren (dejan mashup) heal this empty heart - guiseppe ottaviani shine - john askew fall with me - ben gold, glass child, sneijder one special particle - john O'Callaghan circa forever - rapid eye serenity - gal albutbul Up next week we have local legend Pato De Gomah. Have you joined our FaceBook Group? Sydney Trance Scene Have you joined our FaceBook Page? Sydney Trance Podcast Find us on iTunes! Sydney Trance Itunes Please take a moment to check out our sponsors! Sydney Underground Latest Trance Beatz Radio TMAS TrancenDence

Hey Sydney Trance Fam, It's time for this week's instalment of Sydney Trance. Up this week we have a 2hr special courtesy of Brune aka BB from TrancenDence. This is an absolute cracker of a mix so sit back relax and enjoy. Track list: 1.Ilan Bluestone - Big Ben 2.Above & Beyond Feat. Richard Bedford - On My Way To Mariana Heaven (Above & Beyond Mashup) 3.Jorn Van Deynhoven - New Horizons (A State Of Trance 650 Anthem) (Original Mix) 4.Lost Tribe vs Armin van Buuren & Gaia - Humming Gamemaster (BB Mashup) 5.Armin Van Buuren Pres Gaia - Empire of Hearts (Original Mix) 6.Max Graham - The Evil ID (Mark Sherry Remix) 7.Photographer - Night Lights (Original Mix) 8.Jase Thirlwall - Thunderflash (Indecent Noise Remix) 9.Bowdidge & Taylor - Power Cut (Indecent Noise Remix) 10.Armin van Buuren & Andrew Rayel - Eiforya (Talla 2XLC 140 Remix) 11.Simon Patterson vs John O'Callaghan & Audrey Gallagher - Brush Strokes In The Big Sky (Shura Vlasov Mashup) 12.Darren Porter - Terraforming (Original Mix) 13.Indecent Noise - Warsaw (Original Mix) 14.Pixel & Vini Vici - Anything & Everything (Original mix) 15.Nick Callaghan - Here Today, Gone Tomorrow (Second Sine Remix) 16.Luminary - Amsterdam (Super8 & Tab remix) 17.Max Graham Feat. Ana Criado - Nothing Else Matters (Aly & Fila Remix) 18.Aly & Fila Ft. Susana - Without You (Extended Mix) 19.Aly & Fila - Tula (A & Z Remix) 20.Sean Tyas - Melbourne (Original Mix) Up next week we have local legend Steve Ari. Have you joined our FaceBook Group? Sydney Trance Scene Have you joined our FaceBook Page? Sydney Trance Podcast Find us on iTunes! Sydney Trance Itunes Please take a moment to check out our sponsors! Sydney Underground Latest Trance Beatz Radio TMAS TrancenDence

Hey Sydney Trance Family, It's time for this week's instalment of Sydney Trance. Up this week we have the lovely and extremely talented Amy Parnell. Tune in this week to experience a side of Amy that few of us have had the chance to experience, her PSY Side! Track list: 1)Unseen dimensions, Shake- Wake Up 2) Astrix- Vicious Cycles (Symphonix Remix) 3) Vertical Mode- Bad Luck Buffet 4) Astrix & Dimitri- Evox (Pixel & Freedom Fighters Remix) 5) Omiki- Passion for Music 6) Timelock & Major7- Major Lock 7) Liquid Soul- Adrenaline (Zen Mechanics Remix) 8) John Askew vs John Askew- Giving You Battery Acid (Amy Parnell Mashup) 9) Mekka- Hack The Gibson 10) Coming Soon- What U Mean 11) Ranji, Ziko- Incoming 12)Coming Soon- Become One 13) Captain Monkey- Quemistry Up next week we have BB from TrancenDence with a 2 hr exclusive mix. Have you joined our FaceBook Group? Sydney Trance Scene Have you joined our FaceBook Page? Sydney Trance Podcast Find us on iTunes! Sydney Trance Itunes Please take a moment to check out our sponsors! Sydney Underground Latest Trance Beatz Radio TMAS TrancenDence

Transcript -- At last the TMAs are marked and being returned to the students who are all waiting anxiously to see how they have done.

Transcript -- With their first TMAs submitted its time for the students to relax a little.

With their first TMAs submitted its time for the students to relax a little.

At last the TMAs are marked and being returned to the students who are all waiting anxiously to see how they have done.

A wild episode 41, with Brent Weedman, TMAs, MMAs, Knockouts, Choke outs, but thankfully no flip outs this week! The post Episode 41 – Interview with Brent Weedman appeared first on Hiyaa Martial Arts Podcast.

Comedian/Actor Paul Scheer (NTSF:SD:SUV, Burning Love, The League…), Comedian/Actress June Diane-Raphael (NTSF:SD:SUV, Burning Love, New Girl…), and Writer/Director James Ponsoldt (The Spectacular Now, Smashed, Off the Black...) join us on The Matthew Aaron Show this Wednesday (7/31) as we broadcast LIVE from Taste Chicago in Burbank starting at 4pm PT.   Paul & June took some time to sit down and talk with Matt about their careers and the new 3rd season of NTSF:SD:SUV (currently airing Thursday nights on Adult Swim).   James will be with us live to discuss his Sundance award winning new film THE SPECTACULAR NOW, which stars Shailene Woodley, Mary Elizabeth Winstead & Miles Teller that hits select theaters in NY and LA on August 2nd (expanding wider on August 9th).   Show starts at 4pm PT (6pm CT / 7pm ET). Stop on by Taste Chicago and experience the show in person, otherwise you can stream it from our website or subscribe for free and download the show on iTunes. You can also listen on the go on your Android/iPhone/iPad device via Stitcher. --- This episode is sponsored by · Anchor: The easiest way to make a podcast. https://anchor.fm/app

Actress Judith Hoag (Nashville, Hitchcock, Big Love...) & Aussie husband/wife filmmaking duo Kieran Darcy-Smith & Felicity Price join us on The Matthew Aaron Show Wednesday (6/5) as we broadcast LIVE from Taste Chicago in Burbank starting at 5pm PT. Judith will be will us to discuss her long and diverse career. From originating the role of April O'Neil in TEENAGE MUTANT NINJA TURTLES, to being a series regular on HBO's BIG LOVE and ABC's NASHVILLE, to co-starring in the soon to be released film BAD WORDS (directed by and starring Jason Bateman). Kieran & Felicity were able to take some time to tape an interview earlier with Matt where they discuss their careers and new film WISH YOU WERE HERE. It was written by the duo, directed by Kieran and starring Felicity, Joel Edgerton & Teresa Palmer. You can find it in select US theaters on June 7th. Show starts at 5pm PT (7pm CT / 8pm ET). Stop on by Taste Chicago in Burbank and experience the show in person, otherwise you can stream it from our website or subscribe for free and download the show on iTunes. You can also listen on the go on your Android/iPhone/iPad device via Stitcher. --- This episode is sponsored by · Anchor: The easiest way to make a podcast. https://anchor.fm/app

A special Christmas message, given by Pastor Ted Leavenworth on December 24, 2011.

Believe it... it's Corne from The Most Amazing Show. Also known as Louw Venter. He chats about airline food, nude modelling in the Apartheid era, hair-and-eyebrow acting in adverts, hugging Steve Hofmeyer and the state of the SA entertainment industry.

Season's greetings! Chris and Andrew bring you their review of 2008 and the final part of their evening with Tim. Contact the pair via their new email address of mailbox@beyondtheironsea.com and visit www.beyondtheironsea.com for loads more.

Das Mammakarzinom ist eine Tumorerkrankung mit einem sehr heterogenen Krankheitsverlauf. Um den individuellen Krankheitsverlauf einer Patientin besser vorhersagen zu können und um individuelle Therapieentscheidungen treffen zu können, bedarf es prognostischer und prädiktiver Faktoren. Neben den etablierten Prognosefaktoren Tumorgröße, Lymphknotenstatus, Metastasierung und Grading werden momentan verschiedene biologische Faktoren untersucht. Darüber hinaus wird in letzter Zeit insbesondere die prognostische Bedeutung von disseminierten Tumorzellen im Knochenmark diskutiert. Das Ziel der vorliegenden Arbeit war, die prognostische Relevanz von den biologischen Faktoren HER2, Topoisomerase-IIalpha, Ki67 und p53 sowie deren Korrelation mit dem Nachweis von disseminierten Tumorzellen im Knochenmark bei Patientinnen mit Mamma-CA zu evaluieren. Insgesamt wurden Gewebeproben von 256 primären Mammakarzinomen mit bekanntem Knochenmarkstatus untersucht. Zunächst wurden aus den archivierten Gewebeblöcken sogenannte Tissue-Micro-Arrays (TMAs) hergestellt. An diesen TMAs wurde die Expression von HER2, Topoisomerase-IIalpha, Ki67 und p53 mit Hilfe immunhistochemischer Färbungen analysiert. Zusätzlich wurde das Tumorgewebe auf eine potentielle Genamplifikation von HER2 und Topoisomerase-IIalpha mittels Fluoreszenz-in-situ-Hybridisierung (FISH) hin untersucht. Es stellte sich heraus, dass die meisten dieser biologischen Faktoren untereinander korrelieren. Darüber hinaus konnte eine signifikante Korrelation zwischen der Überexpression von HER2 und einem kürzerem krankheitsfreien Überleben der Patientinnen nachgewiesen werden. Weitere Korrelationen zwischen den einzelnen Faktoren und dem Krankheitsverlauf wurden nur in einigen Subgruppen gefunden. Keiner der untersuchten Faktoren stellte sich jedoch als unabhängiger prognostischer Faktor bezüglich eines kürzeren krankheitsfreien bzw. metastasenfreien Verlaufs oder eines kürzeren Gesamtüberlebens heraus. Diese Ergebnisse deuten daher lediglich auf einen kausalen Zusammenhang zwischen den von mir untersuchten Tumorsuppressormolekülen, Proliferationsmarkern und Wachstumsfaktorrezeptoren hin. Die Expression von HER2, Topoisomerase-IIalpha, Ki67 und p53 auf dem Primärtumor korrelierte nicht mit dem Vorhandensein von Tumorzellen im Knochenmark. Die Dissemination von Tumorzellen ins Knochenmark scheint daher ein von den untersuchten Faktoren unabhängiger Prozess zu sein. Allerdings korrelierte der Nachweis von disseminierten Tumorzellen im Knochenmark mit einem kürzeren Gesamtüberleben der Patientinnen. Bisher gehört die Knochenmarkspunktion zwar nicht zur Routinediagnostik beim primären Mammakarzinom. Allerdings wird diese Untersuchung in einigen Zentren empfohlen. Die weitere Charakterisierung von disseminierten Tumorzellen im Knochenmark sowie die Evaluation von zirkulierenden Tumorzellen im peripheren Blut sind Gegenstand der aktuellen Forschung.

Introduction The prognostic significance of disseminated tumor cells in the bone marrow (DTC-BM) of breast cancer patients has been demonstrated in many studies. Yet, it is not clear which of the primary tumors' biological factors predict hematogenous dissemination. We therefore examined `tissue micro arrays' (TMAs) of 265 primary breast carcinomas from patients with known bone marrow ( BM) status for HER2, Topoisomerase IIa ( Top IIa), Ki 67, and p53. Methods BM analysis was performed by cytospin preparation and immunocytochemical staining for cytokeratin (CK). TMAs were examined by immunohistochemistry (IHC) for HER2, Top IIa, Ki 67 and p53, and fluorescence in situ hybridization ( FISH) for HER2. Results HER2 ( 2+/ 3+) was positive in 35/167 (21%) cases ( FISH 24.3%), Top IIa (> 10%) in 87/187 (46%), Ki 67 in 52/ 184 (28%) and p53 (> 5%) in 61/174 cases (34%). Of 265 patients, 68 (25.7%) showed DTC-BM with a median of 2/2 x 106 cells ( 1 to 1,500). None of the examined factors significantly predicted BM positivity. Significant correlation was seen between HER2 IHC and Top IIa ( p = 0.06), Ki 67 ( p = 0.031), and p53 ( p