Quotes

[clears throat] And also Cursor allows you to submit their MCP server set uh The clears throat And also Cursor allows you to submit their MCP server set uh The clears throat And also Cursor allows you to submit their MCP server set uh The way you submit is different for all three, but basically what All right, so So, when we're serving uh an auto regressive model, these these ch All right, so So, when we're serving uh an auto regressive model, these these ch All right, so So, when we're serving uh an auto regressive model, these these chips are memory bound. Source youtube r305 aQTaU0 — Text Dif And finally, how do you optimize for cost, not just for accuracy, but also for c And finally, how do you optimize for cost, not just for accuracy, but also for c And finally, how do you optimize for cost, not just for accuracy, but also for cost, latency, and reliability before you ship and or as you And for the same amount of evaluation, it takes only 7 hours. And for the same amount of evaluation, it takes only 7 hours. And for the same amount of evaluation, it takes only 7 hours. Source youtube hqHC6Z lXyo — 20 days of compute vs 7 hours rethinking what state of the art mean And lastly, this axis is all about producing more complex work, more representat And lastly, this axis is all about producing more complex work, more representat And lastly, this axis is all about producing more complex work, more representative work, and also nuanced signals that can be used for not And so, to kind of think actually about what a context graph is, we need to ask And so, to kind of think actually about what a context graph is, we need to ask And so, to kind of think actually about what a context graph is, we need to ask ourselves, "Would you agents really be accurate?" Right? Sou And the question is now we need to assign a speaker to each word. And the question is now we need to assign a speaker to each word. And the question is now we need to assign a speaker to each word. Source youtube mFLlVpnGpds — Beyond Transcription Building Voice AI That Understands Con And then context, policies that are um in different reasoning by AI that records And then context, policies that are um in different reasoning by AI that records And then context, policies that are um in different reasoning by AI that records memory, but um by employees and and past humans that have And then we have access to additional tools, memory, or we may need to talk to a And then we have access to additional tools, memory, or we may need to talk to a And then we have access to additional tools, memory, or we may need to talk to additional agents or LLMs like Amazon Nova ACT through somet And this is this is pretty uh important because sometimes the problem is uh, uh And this is this is pretty uh important because sometimes the problem is uh, uh And this is this is pretty uh important because sometimes the problem is uh, uh so critical so we need to fix them as soon as possible and y And today I'm going to be sharing some of what we've learned building voice in v And today I'm going to be sharing some of what we've learned building voice in v And today I'm going to be sharing some of what we've learned building voice in visuals out experiences using AI. Source youtube 65X0pQ6Lmbg And today, I'm going to tell you about our research project, which is called Roa And today, I'm going to tell you about our research project, which is called Roa And today, I'm going to tell you about our research project, which is called Road to 5 million sequence length, breaking memory barriers in And what I Turbo puffer what we think this actually means, you know if you break And what I Turbo puffer what we think this actually means, you know if you break And what I Turbo puffer what we think this actually means, you know if you break down rag into retrieval augmented generation, you know ret But the problem is that state of the art is a bit a confusing concept and people But the problem is that state of the art is a bit a confusing concept and people But the problem is that state of the art is a bit a confusing concept and people maybe have different vision on this. Source youtube hqHC6Z But there's some real breakthroughs over the last few months where both of this But there's some real breakthroughs over the last few months where both of this But there's some real breakthroughs over the last few months where both of this visuals in and audio or the audio in and visuals out experie I'm an engineer here at OpenGov and today we're going to be talking about agents I'm an engineer here at OpenGov and today we're going to be talking about agents I'm an engineer here at OpenGov and today we're going to be talking about agents in production, specifically how OpenGov built and scaled O I'm an open source engineer here at Red Hat, and I think we can all agree that c I'm an open source engineer here at Red Hat, and I think we can all agree that c I'm an open source engineer here at Red Hat, and I think we can all agree that context is the most important aspect to building an AI appli I'm on a mission to solve knowledge representation when all context matters. I'm on a mission to solve knowledge representation when all context matters. I'm on a mission to solve knowledge representation when all context matters. Source youtube XovaGv4f39A — When All Context Matters Extended Cac I'm talking today about text diffusion, which is kind of a more forward-looking I'm talking today about text diffusion, which is kind of a more forward looking I'm talking today about text diffusion, which is kind of a more forward looking research area at DeepMind. Source youtube r305 aQTaU0 — Text If you haven't heard about the Excalidraw MCP app, uh, definitely check it out b If you haven't heard about the Excalidraw MCP app, uh, definitely check it out b If you haven't heard about the Excalidraw MCP app, uh, definitely check it out because the Excalidraw MCP app is not just for um, output an In a world where you are delegating away work to agents and automating away agen In a world where you are delegating away work to agents and automating away agen In a world where you are delegating away work to agents and automating away agents, you need to think about trust boundaries. Source youtub MCP apps is now the official extension of the Model Context Protocol that allows MCP apps is now the official extension of the Model Context Protocol that allows MCP apps is now the official extension of the Model Context Protocol that allows to return UI elements within MCP servers. Source youtube s Model providers you will put evaluation benchmarks on three uh benchmark uh boar Model providers you will put evaluation benchmarks on three uh benchmark uh boar Model providers you will put evaluation benchmarks on three uh benchmark uh boards, but those are not really useful when you put them in yo Most small language models uh for mobile and web are deployed with quantization, Most small language models uh for mobile and web are deployed with quantization, Most small language models uh for mobile and web are deployed with quantization, that is to say, 8 bit, 4 bit, and that can have a quarter Now when the context is provided it's always like these are my security concerns Now when the context is provided it's always like these are my security concerns Now when the context is provided it's always like these are my security concerns which I always have to look into. Source youtube EcqMYoIV5 Of course, Nori ships with everything you need to make this all work, so don't b Of course, Nori ships with everything you need to make this all work, so don't b Of course, Nori ships with everything you need to make this all work, so don't bother reinventing the wheel. Source youtube JRTAtZ5iBkU — H One of the key things is it treat memory as reasoning, not as facts, statistics, One of the key things is it treat memory as reasoning, not as facts, statistics, One of the key things is it treat memory as reasoning, not as facts, statistics, fact with no context and no history, but reasoning. Source Quotes Quotes Transcript backed pull quotes surfaced from AI Engineer World's Fair 2026 supporting videos and livestreams. These are short excerpts selected for navigation and should be checked against the linked transcript/res So for that you need to know who's who who spoke when. So for that you need to know who's who who spoke when. So for that you need to know who's who who spoke when. Source youtube mFLlVpnGpds — Beyond Transcription Building Voice AI That Understands Conversations — Hervé Bre So if you want to build this marketplace of app and have third-party UI rendered So if you want to build this marketplace of app and have third party UI rendered So if you want to build this marketplace of app and have third party UI rendered inside ChatGPT, why not use straight away source doc as th So in there you you'll be able to to with Pianoteq open source toolkit for diari So in there you you'll be able to to with Pianoteq open source toolkit for diari So in there you you'll be able to to with Pianoteq open source toolkit for diarization with piano metrics for evaluation. Source youtube mF So it's not just one pass, it does multiple passes, but it gets to attend to the So it's not just one pass, it does multiple passes, but it gets to attend to the So it's not just one pass, it does multiple passes, but it gets to attend to the future tokens and so on. Source youtube r305 aQTaU0 — Text So you know not not a public benchmark but you can trust the numbers they give u So you know not not a public benchmark but you can trust the numbers they give u So you know not not a public benchmark but you can trust the numbers they give us. Source youtube UM6sFg jdlE — RAG is dead, right?? — Kuba So, basically, this will lead to context overload and uh we need to manage that So, basically, this will lead to context overload and uh we need to manage that So, basically, this will lead to context overload and uh we need to manage that properly. Source youtube vh2VGuQ3zhY — The 100 Tool Agent Is So, let's see, uh, a what we have in our the current system uh, and current memo So, let's see, uh, a what we have in our the current system uh, and current memo So, let's see, uh, a what we have in our the current system uh, and current memory is that they basically store user preferences, profile, So, that's kind of how we handled long context and memory, and it's worked prett So, that's kind of how we handled long context and memory, and it's worked prett So, that's kind of how we handled long context and memory, and it's worked pretty well for us. Source youtube 4uFVSLgD2Q4 — Agents in Produ So, uh just wanted to share a little bit about that and how we've uh solved the So, uh just wanted to share a little bit about that and how we've uh solved the So, uh just wanted to share a little bit about that and how we've uh solved the long context problem. Source youtube 4uFVSLgD2Q4 — Agents in That context pointer literally just says if you need the template or if you need That context pointer literally just says if you need the template or if you need That context pointer literally just says if you need the template or if you need to update the context.md file, go to this file. Source you The event outcome becomes a first-class signal in the retrieval re-ranking and n The event outcome becomes a first class signal in the retrieval re ranking and n The event outcome becomes a first class signal in the retrieval re ranking and not just for retrieval. Source youtube Jx4ZFEAq6bY — User Si The human still decides what context matters, still remembers what to ask, still The human still decides what context matters, still remembers what to ask, still The human still decides what context matters, still remembers what to ask, still chooses the timing, still notices the ambiguity, still rep The let's say past few decades, we have been building the web for human actions The let's say past few decades, we have been building the web for human actions The let's say past few decades, we have been building the web for human actions and human eyes, and we've been trying to optimize for that, The problem here is that the context window is limited, and if you fill the cont The problem here is that the context window is limited, and if you fill the cont The problem here is that the context window is limited, and if you fill the context window too much, the quality of the answer gets degrade This approach would look something like \"cache augmented generation\" (CAG), wher This approach would look something like "cache augmented generation" CAG , wher This approach would look something like "cache augmented generation" CAG , where we use a model with a large context window, load the docume This continual learning loop needs to run frequently and we need to have efficie This continual learning loop needs to run frequently and we need to have efficie This continual learning loop needs to run frequently and we need to have efficiency in uh different layers in updates to the agent. Source Uh as you continue scaling your context, your memory keeps growing linearly, whi Uh as you continue scaling your context, your memory keeps growing linearly, whi Uh as you continue scaling your context, your memory keeps growing linearly, which is not as bad, but still pretty difficult to deal with, Uh I had a question on um on how how much evaluation you need to write for uh fe Uh I had a question on um on how how much evaluation you need to write for uh fe Uh I had a question on um on how how much evaluation you need to write for uh feature cuz especially when you run against live traces, some Uh it's an open source inspector, again, uh something comparable to the official Uh it's an open source inspector, again, uh something comparable to the official Uh it's an open source inspector, again, uh something comparable to the official inspector from the model context protocol maintainers that Uh the other thing to call out here is like we already have a lot of companies u Uh the other thing to call out here is like we already have a lot of companies u Uh the other thing to call out here is like we already have a lot of companies using brain trust in production today. Source youtube bk0Tmx Uh when you get uh into the real meat of evaluation is when you do custom code e Uh when you get uh into the real meat of evaluation is when you do custom code e Uh when you get uh into the real meat of evaluation is when you do custom code evals. Source youtube Xfl50508LZM — Ship Real Agents Hands O Uh, so, you need to know which of your evals is which, which ones should be ship Uh, so, you need to know which of your evals is which, which ones should be ship Uh, so, you need to know which of your evals is which, which ones should be ship blockers, and which ones are just informative. Source yout Um but essentially an MCP app is going to allow us to really interact or stay wi Um but essentially an MCP app is going to allow us to really interact or stay wi Um but essentially an MCP app is going to allow us to really interact or stay within one context such as VS Code. Source youtube xIwFcnHqp4 Um But yeah, that's a good point and then in the create context graph, we're sti Um But yeah, that's a good point and then in the create context graph, we're sti Um But yeah, that's a good point and then in the create context graph, we're still working on how you would write um new decision traces. S Um so anyway, so the the title of my talk today is evals are broken and you shou Um so anyway, so the the title of my talk today is evals are broken and you shou Um so anyway, so the the title of my talk today is evals are broken and you should use them anyway. Source youtube QuuIywMG4s8 — Evals Are Um So what we're finding now is that a lot of people are no longer doing the sim Um So what we're finding now is that a lot of people are no longer doing the sim Um So what we're finding now is that a lot of people are no longer doing the simple rag you know the the Twitter quote unquote rag of just Um to do that all effectively, you need to make sure that the models are able to Um to do that all effectively, you need to make sure that the models are able to Um to do that all effectively, you need to make sure that the models are able to process that context and work with it correctly at the tra Um you need to before you start doing anything, you need to know what it is that Um you need to before you start doing anything, you need to know what it is that Um you need to before you start doing anything, you need to know what it is that you're measuring. Source youtube fWXJM J0ZB8 — Frontier re We can just switch to having voice in visuals out and then we benefit from the m We can just switch to having voice in visuals out and then we benefit from the m We can just switch to having voice in visuals out and then we benefit from the more forgiving visual response envelope that people have. So We need to think about the branches in our skill moving material out behind cont We need to think about the branches in our skill moving material out behind cont We need to think about the branches in our skill moving material out behind context pointers. Source youtube UNzCG3lw6O0 — Building Great A Well, today I'm going to share four engineering lessons from the Chrome uh DevTo Well, today I'm going to share four engineering lessons from the Chrome uh DevTo Well, today I'm going to share four engineering lessons from the Chrome uh DevTools team on how we build Chrome DevTools for agents and how When you're building agents, not just using them to write code, you start gettin When you're building agents, not just using them to write code, you start gettin When you're building agents, not just using them to write code, you start getting into architecting agentic systems. Source youtube ZD9 4fW