[
  {
    "startMs": 1309,
    "text": "[music]"
  },
  {
    "startMs": 13040,
    "text": ">> Hello everyone."
  },
  {
    "startMs": 14400,
    "text": "Um"
  },
  {
    "startMs": 15800,
    "text": "So, I want to start with a simple"
  },
  {
    "startMs": 17920,
    "text": "question."
  },
  {
    "startMs": 19160,
    "text": "What if your team or your org or company"
  },
  {
    "startMs": 22520,
    "text": "moves like a single body?"
  },
  {
    "startMs": 24360,
    "text": "I'm Abdullah Muhammad, the VP of AIML at"
  },
  {
    "startMs": 26920,
    "text": "A Data Chef."
  },
  {
    "startMs": 28120,
    "text": "And today was supposed to be Khalid with"
  },
  {
    "startMs": 29960,
    "text": "me to present this, but he's he's down"
  },
  {
    "startMs": 32080,
    "text": "with our development partner at the"
  },
  {
    "startMs": 33400,
    "text": "moment."
  },
  {
    "startMs": 34800,
    "text": "So, I will be presenting the whole"
  },
  {
    "startMs": 36200,
    "text": "presentation for today."
  },
  {
    "startMs": 38160,
    "text": "So, let's go for the next slide."
  },
  {
    "startMs": 41280,
    "text": "So, how many of you have been attending"
  },
  {
    "startMs": 43520,
    "text": "the World Cup soccer or watching some"
  },
  {
    "startMs": 46440,
    "text": "games on ice? We have a couple of fans."
  },
  {
    "startMs": 49240,
    "text": "Yeah, it's over all over the place. And"
  },
  {
    "startMs": 51400,
    "text": "imagine for a moment, just a single"
  },
  {
    "startMs": 53720,
    "text": "moment, you are a soccer player, all"
  },
  {
    "startMs": 56320,
    "text": "right?"
  },
  {
    "startMs": 58000,
    "text": "And"
  },
  {
    "startMs": 59320,
    "text": "if you are a soccer player, you have"
  },
  {
    "startMs": 60680,
    "text": "this intent the moment you go into the"
  },
  {
    "startMs": 63360,
    "text": "field, you just going to run and score a"
  },
  {
    "startMs": 66920,
    "text": "goal."
  },
  {
    "startMs": 67840,
    "text": "This is what you want to do."
  },
  {
    "startMs": 70160,
    "text": "And for the second thing, you have this"
  },
  {
    "startMs": 72800,
    "text": "knowledge that you've been accumulated"
  },
  {
    "startMs": 74640,
    "text": "through your training the whole day,"
  },
  {
    "startMs": 76800,
    "text": "your exercises with your coach, and the"
  },
  {
    "startMs": 79120,
    "text": "best practices and the videos you have"
  },
  {
    "startMs": 80880,
    "text": "watched."
  },
  {
    "startMs": 81920,
    "text": "And you at the moment in the field like"
  },
  {
    "startMs": 84040,
    "text": "the moment of truth that you are there,"
  },
  {
    "startMs": 85520,
    "text": "you combine both of the intent and"
  },
  {
    "startMs": 87520,
    "text": "knowledge"
  },
  {
    "startMs": 88720,
    "text": "and compound both of them."
  },
  {
    "startMs": 90480,
    "text": "And through your nervous system, you"
  },
  {
    "startMs": 92120,
    "text": "execute to achieve your goal."
  },
  {
    "startMs": 95000,
    "text": "And we can call this, in a sense,"
  },
  {
    "startMs": 98000,
    "text": "you are being self-aligned as a single"
  },
  {
    "startMs": 100920,
    "text": "entity by yourself."
  },
  {
    "startMs": 104400,
    "text": "And accept the fact that a soccer team"
  },
  {
    "startMs": 108040,
    "text": "or a football team, depending where"
  },
  {
    "startMs": 109520,
    "text": "you're coming from, is not a single"
  },
  {
    "startMs": 111280,
    "text": "player. It's actually"
  },
  {
    "startMs": 113080,
    "text": "11 players."
  },
  {
    "startMs": 114720,
    "text": "And on the field, you are up against"
  },
  {
    "startMs": 117120,
    "text": "another team with 11 players they"
  },
  {
    "startMs": 119720,
    "text": "playing against you. And at this moment"
  },
  {
    "startMs": 124160,
    "text": "it's not about your individual skills,"
  },
  {
    "startMs": 127040,
    "text": "it's about how your team working"
  },
  {
    "startMs": 128720,
    "text": "together well."
  },
  {
    "startMs": 131000,
    "text": "So,"
  },
  {
    "startMs": 132520,
    "text": "in general"
  },
  {
    "startMs": 134000,
    "text": "like the team keep changing and"
  },
  {
    "startMs": 136080,
    "text": "everything is getting harder and harder"
  },
  {
    "startMs": 138120,
    "text": "and the team that wins actually the team"
  },
  {
    "startMs": 140480,
    "text": "that the most aligned"
  },
  {
    "startMs": 142520,
    "text": "in both of the both of teams. So, in"
  },
  {
    "startMs": 145400,
    "text": "short we can say alignment beats"
  },
  {
    "startMs": 148640,
    "text": "individual skills."
  },
  {
    "startMs": 151306,
    "text": ">> [snorts]"
  },
  {
    "startMs": 152040,
    "text": ">> Okay."
  },
  {
    "startMs": 153480,
    "text": "Now, what if your team is over 50"
  },
  {
    "startMs": 156280,
    "text": "engineers or 50 players?"
  },
  {
    "startMs": 158800,
    "text": "This is completely changes the whole"
  },
  {
    "startMs": 160880,
    "text": "scene right now."
  },
  {
    "startMs": 162400,
    "text": "So, everyone at these days"
  },
  {
    "startMs": 165600,
    "text": "we empower the engineers with AI tools,"
  },
  {
    "startMs": 168880,
    "text": "AI agents and we want to increase the"
  },
  {
    "startMs": 170920,
    "text": "productivity."
  },
  {
    "startMs": 172400,
    "text": "But we know from literature that the"
  },
  {
    "startMs": 174480,
    "text": "more people you have, the quadratic term"
  },
  {
    "startMs": 177080,
    "text": "of communication between them and"
  },
  {
    "startMs": 178560,
    "text": "alignment them keep growing and keep"
  },
  {
    "startMs": 180800,
    "text": "growing. And at a specific point"
  },
  {
    "startMs": 182840,
    "text": "actually"
  },
  {
    "startMs": 185440,
    "text": "it actually starts to going declining."
  },
  {
    "startMs": 187840,
    "text": "Your throughput actually is not what you"
  },
  {
    "startMs": 189400,
    "text": "getting."
  },
  {
    "startMs": 190480,
    "text": "It's diminishing cost."
  },
  {
    "startMs": 192200,
    "text": "So, everyone trying to solve this linear"
  },
  {
    "startMs": 194840,
    "text": "problem of more tools and more stuff but"
  },
  {
    "startMs": 196840,
    "text": "nobody actually tackling the quadratic"
  },
  {
    "startMs": 198600,
    "text": "term over there."
  },
  {
    "startMs": 200040,
    "text": "And this is why the alignment is"
  },
  {
    "startMs": 201360,
    "text": "important. If you are able to change"
  },
  {
    "startMs": 202840,
    "text": "this quadratic term into a linear term"
  },
  {
    "startMs": 205320,
    "text": "or build a multi-layer AI system that"
  },
  {
    "startMs": 208640,
    "text": "will solve this problem."
  },
  {
    "startMs": 212880,
    "text": "Okay, moving into ship design. Ship"
  },
  {
    "startMs": 215239,
    "text": "design is"
  },
  {
    "startMs": 217480,
    "text": "a different story."
  },
  {
    "startMs": 219160,
    "text": "If you are in software company, you have"
  },
  {
    "startMs": 221320,
    "text": "a bug in your software, you can ship a"
  },
  {
    "startMs": 223640,
    "text": "batch to fix it. You can roll out a new"
  },
  {
    "startMs": 226200,
    "text": "version. It's most of the time is"
  },
  {
    "startMs": 228080,
    "text": "doable."
  },
  {
    "startMs": 229239,
    "text": "But in ships, you can't do this in"
  },
  {
    "startMs": 230720,
    "text": "ships. It's hardware, fixed it on"
  },
  {
    "startMs": 232640,
    "text": "silicon has been printed."
  },
  {
    "startMs": 234560,
    "text": "And if you're going to do this, there is"
  },
  {
    "startMs": 236840,
    "text": "a cost actually, we call it the risk"
  },
  {
    "startMs": 238400,
    "text": "band cost."
  },
  {
    "startMs": 240080,
    "text": "On average between chip design companies"
  },
  {
    "startMs": 243400,
    "text": "about $50 million."
  },
  {
    "startMs": 245720,
    "text": "And for some companies like being 1"
  },
  {
    "startMs": 248400,
    "text": "month late in the market, it's a make or"
  },
  {
    "startMs": 250320,
    "text": "break for them."
  },
  {
    "startMs": 252200,
    "text": "And we spoke to many practitioners in"
  },
  {
    "startMs": 253959,
    "text": "the field. On average like 15"
  },
  {
    "startMs": 255840,
    "text": "practitioner and we found that most of"
  },
  {
    "startMs": 258040,
    "text": "them pointed towards the same problem."
  },
  {
    "startMs": 260480,
    "text": "That we spend 70% of our time doing"
  },
  {
    "startMs": 264640,
    "text": "alignment."
  },
  {
    "startMs": 266520,
    "text": "Alignment to make sure that once we"
  },
  {
    "startMs": 268240,
    "text": "print the chip, nothing is there. And"
  },
  {
    "startMs": 270440,
    "text": "one of the key words that we heard and"
  },
  {
    "startMs": 273080,
    "text": "still resonating that the most"
  },
  {
    "startMs": 275040,
    "text": "successful chip organization are not the"
  },
  {
    "startMs": 277600,
    "text": "one with the best engineers,"
  },
  {
    "startMs": 279720,
    "text": "but they are the most aligned organized."
  },
  {
    "startMs": 284440,
    "text": "So, how chip design today works?"
  },
  {
    "startMs": 289560,
    "text": "We start with bottom figure like the"
  },
  {
    "startMs": 291960,
    "text": "fragmented intent and decision. You"
  },
  {
    "startMs": 293640,
    "text": "attend couple of meetings, you talk"
  },
  {
    "startMs": 295520,
    "text": "about decisions, what you're going to do"
  },
  {
    "startMs": 297040,
    "text": "next. You have the specs written"
  },
  {
    "startMs": 299160,
    "text": "everywhere, you have the Slack messages,"
  },
  {
    "startMs": 301160,
    "text": "you have emails, everything is"
  },
  {
    "startMs": 303000,
    "text": "fragmented over there."
  },
  {
    "startMs": 304560,
    "text": "And then we go into a second part, which"
  },
  {
    "startMs": 307240,
    "text": "is the knowledge."
  },
  {
    "startMs": 308680,
    "text": "Nobody updates wikis, right? Many of us"
  },
  {
    "startMs": 310960,
    "text": "has wikis."
  },
  {
    "startMs": 312280,
    "text": "They've been collecting dust for years"
  },
  {
    "startMs": 314360,
    "text": "and the code keep evolving outside the"
  },
  {
    "startMs": 316280,
    "text": "wikis, it's not over there."
  },
  {
    "startMs": 318560,
    "text": "And now we have the tools that you"
  },
  {
    "startMs": 320680,
    "text": "execute with, which comes with many many"
  },
  {
    "startMs": 322880,
    "text": "fractions."
  },
  {
    "startMs": 324200,
    "text": "And these tools like the data is lost"
  },
  {
    "startMs": 326480,
    "text": "over there, what input, what output,"
  },
  {
    "startMs": 328160,
    "text": "what results, most of the time are not"
  },
  {
    "startMs": 330040,
    "text": "being captured."
  },
  {
    "startMs": 331560,
    "text": "And what you see here is not something"
  },
  {
    "startMs": 333200,
    "text": "we came with like draw from our imagine,"
  },
  {
    "startMs": 336520,
    "text": "this is actually how is it today. We"
  },
  {
    "startMs": 338200,
    "text": "wrote from inside the companies and from"
  },
  {
    "startMs": 340800,
    "text": "the backgrounds of the people we have in"
  },
  {
    "startMs": 342240,
    "text": "our team."
  },
  {
    "startMs": 344560,
    "text": "And what we're trying to solve here is"
  },
  {
    "startMs": 346480,
    "text": "building a multi-layer AI"
  },
  {
    "startMs": 348680,
    "text": "with a shared nervous system. Instead of"
  },
  {
    "startMs": 351240,
    "text": "having"
  },
  {
    "startMs": 352360,
    "text": "scattered knowledge or scattered intent"
  },
  {
    "startMs": 355680,
    "text": "all over the place, we build a living"
  },
  {
    "startMs": 358000,
    "text": "graph. We call it the system of intent."
  },
  {
    "startMs": 360760,
    "text": "And this living graph actually"
  },
  {
    "startMs": 363920,
    "text": "has all the constraints of the system,"
  },
  {
    "startMs": 366400,
    "text": "has all the decisions over there. It"
  },
  {
    "startMs": 368640,
    "text": "keep evolving. And as an AI person"
  },
  {
    "startMs": 372120,
    "text": "actually,"
  },
  {
    "startMs": 373160,
    "text": "we don't allow the agents to touch it"
  },
  {
    "startMs": 375400,
    "text": "except with human in the loop approval"
  },
  {
    "startMs": 378040,
    "text": "for specific changes. And this thing is"
  },
  {
    "startMs": 380440,
    "text": "like the Bible of the whole system. This"
  },
  {
    "startMs": 382200,
    "text": "is where the whole org is going or whole"
  },
  {
    "startMs": 384960,
    "text": "company is going."
  },
  {
    "startMs": 386680,
    "text": "And the next one is the tribal knowledge"
  },
  {
    "startMs": 388400,
    "text": "layer. The tribal knowledge layer"
  },
  {
    "startMs": 390520,
    "text": "we can think about it as a memory"
  },
  {
    "startMs": 392760,
    "text": "that keeps evolving with day-to-day"
  },
  {
    "startMs": 394919,
    "text": "usage and the knowledge base that"
  },
  {
    "startMs": 396760,
    "text": "capture all the information and"
  },
  {
    "startMs": 398200,
    "text": "documents. And it's keep evolving from a"
  },
  {
    "startMs": 400680,
    "text": "project to project and keeping the best"
  },
  {
    "startMs": 403280,
    "text": "practice over there."
  },
  {
    "startMs": 405280,
    "text": "And lastly,"
  },
  {
    "startMs": 406800,
    "text": "instead of having this general coding"
  },
  {
    "startMs": 409520,
    "text": "agent that everyone uses today, we have"
  },
  {
    "startMs": 411919,
    "text": "a special designed agent that being"
  },
  {
    "startMs": 414440,
    "text": "developed by subject matter experts"
  },
  {
    "startMs": 416880,
    "text": "to help the engineers doing their work."
  },
  {
    "startMs": 420600,
    "text": "So for example, like we have"
  },
  {
    "startMs": 422640,
    "text": "digital design agent, analog design"
  },
  {
    "startMs": 425800,
    "text": "agent, and so on."
  },
  {
    "startMs": 428040,
    "text": "And by combining all of this, you will"
  },
  {
    "startMs": 429800,
    "text": "have this shared nervous system that"
  },
  {
    "startMs": 432040,
    "text": "allows you to move fast and move"
  },
  {
    "startMs": 433680,
    "text": "forward."
  },
  {
    "startMs": 438040,
    "text": "Okay. So it's easy to say an idea on a"
  },
  {
    "startMs": 440560,
    "text": "slide. It's nice. Everyone makes slides."
  },
  {
    "startMs": 443440,
    "text": "But I want to show you like a demo from"
  },
  {
    "startMs": 445680,
    "text": "what we have today"
  },
  {
    "startMs": 447400,
    "text": "and showing the intent, knowledge, and"
  },
  {
    "startMs": 449000,
    "text": "execution. It will be short demos."
  },
  {
    "startMs": 452080,
    "text": "And we'll start with the first one."
  },
  {
    "startMs": 455520,
    "text": "Yeah, the"
  },
  {
    "startMs": 456720,
    "text": "Yeah. Okay, cool. So we can see that"
  },
  {
    "startMs": 459560,
    "text": "each engineer gets a role-based AI"
  },
  {
    "startMs": 462360,
    "text": "teammate specific to their role."
  },
  {
    "startMs": 465120,
    "text": "They can check the knowledge base of the"
  },
  {
    "startMs": 466680,
    "text": "whole project that being contained and"
  },
  {
    "startMs": 469000,
    "text": "being growing and compounding over time"
  },
  {
    "startMs": 471280,
    "text": "and now they have their own intent."
  },
  {
    "startMs": 474800,
    "text": "And"
  },
  {
    "startMs": 476840,
    "text": "you have single place"
  },
  {
    "startMs": 478840,
    "text": "for design"
  },
  {
    "startMs": 480680,
    "text": "where it captures all"
  },
  {
    "startMs": 483600,
    "text": "the tooling you have. It captures the"
  },
  {
    "startMs": 485760,
    "text": "results. It captures what you did and"
  },
  {
    "startMs": 488120,
    "text": "what you're going to do next and"
  },
  {
    "startMs": 489320,
    "text": "analysis of everything. So, everything"
  },
  {
    "startMs": 491600,
    "text": "being contained in one place."
  },
  {
    "startMs": 497080,
    "text": "And here we see"
  },
  {
    "startMs": 500640,
    "text": "a human finishing their work."
  },
  {
    "startMs": 503360,
    "text": "This human signing off the"
  },
  {
    "startMs": 506240,
    "text": "the results of some space simulation"
  },
  {
    "startMs": 508880,
    "text": "and the system of intent realizes, okay,"
  },
  {
    "startMs": 510880,
    "text": "this person is done with this. I'm going"
  },
  {
    "startMs": 512919,
    "text": "to notify the next stakeholders of what"
  },
  {
    "startMs": 515159,
    "text": "they should do"
  },
  {
    "startMs": 516400,
    "text": "and signal to them that they are done"
  },
  {
    "startMs": 518840,
    "text": "with this."
  },
  {
    "startMs": 522680,
    "text": "And now the system of intent, which"
  },
  {
    "startMs": 524280,
    "text": "actually the nervous system or the Bible"
  },
  {
    "startMs": 526800,
    "text": "of the system, it's a graph living graph"
  },
  {
    "startMs": 528680,
    "text": "that keep compounding with time. We see"
  },
  {
    "startMs": 530600,
    "text": "in this example, like it realizes like"
  },
  {
    "startMs": 533240,
    "text": "there is something off, like some value"
  },
  {
    "startMs": 535760,
    "text": "out of constraints that shouldn't be"
  },
  {
    "startMs": 537360,
    "text": "there that might cost you $50 million"
  },
  {
    "startMs": 539640,
    "text": "actually to suspend the whole ship"
  },
  {
    "startMs": 541640,
    "text": "and it notified the system and the"
  },
  {
    "startMs": 543520,
    "text": "notification goes and some engineers"
  },
  {
    "startMs": 545760,
    "text": "start working on it. And once it got it"
  },
  {
    "startMs": 548320,
    "text": "fixed, it submit a game into the system"
  },
  {
    "startMs": 550960,
    "text": "and it keep evolving over time."
  },
  {
    "startMs": 555280,
    "text": "Okay."
  },
  {
    "startMs": 556640,
    "text": "Good."
  },
  {
    "startMs": 558680,
    "text": "So, let's say for example, like um"
  },
  {
    "startMs": 562440,
    "text": "you were working in the system, you look"
  },
  {
    "startMs": 564920,
    "text": "at the Bible, you find, oh, there"
  },
  {
    "startMs": 567120,
    "text": "there's something wrong about it. Uh I"
  },
  {
    "startMs": 569480,
    "text": "don't like this value."
  },
  {
    "startMs": 571200,
    "text": "And then you propose a change. So, the"
  },
  {
    "startMs": 573440,
    "text": "system of intent and this big graph"
  },
  {
    "startMs": 576080,
    "text": "captures all the values over there, all"
  },
  {
    "startMs": 578960,
    "text": "the stakeholders, and you start doing"
  },
  {
    "startMs": 580600,
    "text": "this modification,"
  },
  {
    "startMs": 582280,
    "text": "and it gather all the shared knowledge,"
  },
  {
    "startMs": 586120,
    "text": "and then it fire a request, as you can"
  },
  {
    "startMs": 589000,
    "text": "see here,"
  },
  {
    "startMs": 590200,
    "text": "and this request goes to an architect or"
  },
  {
    "startMs": 592400,
    "text": "an owner of the system. The owner can"
  },
  {
    "startMs": 594440,
    "text": "approve or decline it, and the moment"
  },
  {
    "startMs": 596040,
    "text": "they approve that this is a valid"
  },
  {
    "startMs": 598280,
    "text": "change, it actually goes and echo in the"
  },
  {
    "startMs": 600320,
    "text": "whole system. Like everyone will know"
  },
  {
    "startMs": 602600,
    "text": "that this decision has been made. There"
  },
  {
    "startMs": 604640,
    "text": "is that change that advises everything"
  },
  {
    "startMs": 606840,
    "text": "over there."
  },
  {
    "startMs": 608880,
    "text": "What?"
  },
  {
    "startMs": 614440,
    "text": "Good. So,"
  },
  {
    "startMs": 617920,
    "text": "moving to a very difficult topic we have"
  },
  {
    "startMs": 619880,
    "text": "like how we going to evaluate uh our"
  },
  {
    "startMs": 623000,
    "text": "claims and measure the success of the"
  },
  {
    "startMs": 625680,
    "text": "system."
  },
  {
    "startMs": 628520,
    "text": "The philosophy we are"
  },
  {
    "startMs": 630120,
    "text": "using this or the philosophy toward"
  },
  {
    "startMs": 632280,
    "text": "this, we don't grade the agents."
  },
  {
    "startMs": 634520,
    "text": "We try to grade alignment itself. So, we"
  },
  {
    "startMs": 636800,
    "text": "have four axes, two horizontal, two"
  },
  {
    "startMs": 639320,
    "text": "vertical. The horizontal axes like"
  },
  {
    "startMs": 641360,
    "text": "qualitative,"
  },
  {
    "startMs": 643200,
    "text": "the vertical axes like qualitative and"
  },
  {
    "startMs": 644920,
    "text": "quantitative values, which is typical in"
  },
  {
    "startMs": 647680,
    "text": "this domain at the moment. And then"
  },
  {
    "startMs": 649440,
    "text": "horizontal ones, which is bare component"
  },
  {
    "startMs": 651520,
    "text": "and the system into it."
  },
  {
    "startMs": 653520,
    "text": "And [snorts] if we're going to zoom into"
  },
  {
    "startMs": 655160,
    "text": "the bare component, you can measure like"
  },
  {
    "startMs": 657520,
    "text": "if that agent"
  },
  {
    "startMs": 659920,
    "text": "giving you the correct output for this"
  },
  {
    "startMs": 661960,
    "text": "voltage, like known values versus golden"
  },
  {
    "startMs": 664360,
    "text": "answers. Or you can use LNM judge and"
  },
  {
    "startMs": 667240,
    "text": "measure the golden answer versus the"
  },
  {
    "startMs": 668839,
    "text": "expert we have for this one. Which is"
  },
  {
    "startMs": 671880,
    "text": "okay. You can measure how good my"
  },
  {
    "startMs": 674320,
    "text": "memory, like if the recall state of art,"
  },
  {
    "startMs": 676640,
    "text": "which is the case in our thing. Are we"
  },
  {
    "startMs": 678760,
    "text": "doing inference really good?"
  },
  {
    "startMs": 681000,
    "text": "But then it comes into the harder"
  },
  {
    "startMs": 682440,
    "text": "question, which is basically are we"
  },
  {
    "startMs": 685640,
    "text": "doing a task completion? Like if someone"
  },
  {
    "startMs": 687560,
    "text": "uses this whole thing,"
  },
  {
    "startMs": 689440,
    "text": "is he really completing the task he want"
  },
  {
    "startMs": 691960,
    "text": "to do."
  },
  {
    "startMs": 693040,
    "text": "Is he frustrated while using this? Are"
  },
  {
    "startMs": 695400,
    "text": "our agent overstepping human in the loop"
  },
  {
    "startMs": 697760,
    "text": "approval or not?"
  },
  {
    "startMs": 699120,
    "text": "Sometimes the agent go goes out on that"
  },
  {
    "startMs": 701240,
    "text": "end."
  },
  {
    "startMs": 702400,
    "text": "And we measure also does does our system"
  },
  {
    "startMs": 706240,
    "text": "allow you to work concurrently on"
  },
  {
    "startMs": 708240,
    "text": "multiple task in parallel? This is a"
  },
  {
    "startMs": 710400,
    "text": "success metric or success goal we have."
  },
  {
    "startMs": 713240,
    "text": "And the last one is token tax. We don't"
  },
  {
    "startMs": 715240,
    "text": "want to overload you once you use this"
  },
  {
    "startMs": 717200,
    "text": "with all the lovely tokens and increase"
  },
  {
    "startMs": 719840,
    "text": "your budget."
  },
  {
    "startMs": 721560,
    "text": "And there is hard frontier here like in"
  },
  {
    "startMs": 724520,
    "text": "the future now the topic of memory or"
  },
  {
    "startMs": 727400,
    "text": "graph memory or graph rag, whatever the"
  },
  {
    "startMs": 730000,
    "text": "title is,"
  },
  {
    "startMs": 731200,
    "text": "is there is around like 150 papers in"
  },
  {
    "startMs": 733840,
    "text": "this area at the moment and all of them"
  },
  {
    "startMs": 736080,
    "text": "are addressing in a nice way. You can"
  },
  {
    "startMs": 737680,
    "text": "measure the recall there is data sets."
  },
  {
    "startMs": 740080,
    "text": "But there is no"
  },
  {
    "startMs": 741960,
    "text": "work and research at the moment that"
  },
  {
    "startMs": 744440,
    "text": "targets tribal memory or institutional"
  },
  {
    "startMs": 746680,
    "text": "memory. Like what does it mean exactly?"
  },
  {
    "startMs": 748880,
    "text": "How do you measure tribal memory"
  },
  {
    "startMs": 751120,
    "text": "success?"
  },
  {
    "startMs": 752280,
    "text": "And also for the chip design domain,"
  },
  {
    "startMs": 754120,
    "text": "it's actually even harder because there"
  },
  {
    "startMs": 756480,
    "text": "is not enough data sets like computer"
  },
  {
    "startMs": 758560,
    "text": "vision domain, there is many data sets"
  },
  {
    "startMs": 760320,
    "text": "over there. So there is nothing"
  },
  {
    "startMs": 761920,
    "text": "collected. So we have our own wheel and"
  },
  {
    "startMs": 764480,
    "text": "going with SMEs collecting this kind of"
  },
  {
    "startMs": 767320,
    "text": "data sets."
  },
  {
    "startMs": 769680,
    "text": "Cool."
  },
  {
    "startMs": 770760,
    "text": "So [snorts] what broke? Which actually"
  },
  {
    "startMs": 773160,
    "text": "when I attend any talk I like to hear"
  },
  {
    "startMs": 774880,
    "text": "what broke, how do you fix it?"
  },
  {
    "startMs": 777440,
    "text": "First, agent overstepped."
  },
  {
    "startMs": 780000,
    "text": "In early design phases of the system, we"
  },
  {
    "startMs": 782040,
    "text": "found that an analog agent that's"
  },
  {
    "startMs": 785080,
    "text": "specifically for analog design actually"
  },
  {
    "startMs": 787640,
    "text": "overstepping and doing RTL agent work."
  },
  {
    "startMs": 791000,
    "text": "Which wasn't really great."
  },
  {
    "startMs": 793120,
    "text": "Even we tried to enforce it, but it was"
  },
  {
    "startMs": 794960,
    "text": "a difficult problem. And then another"
  },
  {
    "startMs": 797000,
    "text": "thing is we noticed that truth has"
  },
  {
    "startMs": 799839,
    "text": "drifted. An agent modifying something in"
  },
  {
    "startMs": 802520,
    "text": "the system"
  },
  {
    "startMs": 804000,
    "text": "not necessarily means it modifies it"
  },
  {
    "startMs": 806000,
    "text": "everywhere it should be modified."
  },
  {
    "startMs": 808920,
    "text": "And that make it harder. Like we have"
  },
  {
    "startMs": 810360,
    "text": "the cases specifically where one agent"
  },
  {
    "startMs": 812800,
    "text": "were modifying a parameter, it updated"
  },
  {
    "startMs": 814560,
    "text": "it in one place, five other places were"
  },
  {
    "startMs": 816960,
    "text": "forgotten."
  },
  {
    "startMs": 818800,
    "text": "And the third one is one of my favorite"
  },
  {
    "startMs": 821040,
    "text": "is we asked the agent do not write into"
  },
  {
    "startMs": 823800,
    "text": "specs. Just don't don't change the"
  },
  {
    "startMs": 825640,
    "text": "specs."
  },
  {
    "startMs": 826840,
    "text": "They said, \"Okay, I obey you. I'm not"
  },
  {
    "startMs": 829120,
    "text": "going to write into specs.\" But then"
  },
  {
    "startMs": 831440,
    "text": "they moved into bash and they used set"
  },
  {
    "startMs": 834240,
    "text": "to write into specs. We blocked, bash we"
  },
  {
    "startMs": 836560,
    "text": "blocked set. They said, \"Okay, cool. I"
  },
  {
    "startMs": 838720,
    "text": "will use cat actually to write over the"
  },
  {
    "startMs": 840800,
    "text": "specs.\" So we're being like"
  },
  {
    "startMs": 843040,
    "text": "a cat chasing a mouse around to just to"
  },
  {
    "startMs": 845080,
    "text": "prevent it from writing over specs."
  },
  {
    "startMs": 847800,
    "text": "And based on these three failures we"
  },
  {
    "startMs": 849680,
    "text": "have,"
  },
  {
    "startMs": 850880,
    "text": "we came up with principles that we are"
  },
  {
    "startMs": 853160,
    "text": "working today."
  },
  {
    "startMs": 854960,
    "text": "First, we have a spec hierarchy with"
  },
  {
    "startMs": 858000,
    "text": "agent scope and file isolation to allow"
  },
  {
    "startMs": 860680,
    "text": "them only to work on this specific task"
  },
  {
    "startMs": 862800,
    "text": "or specific domain. That's all this"
  },
  {
    "startMs": 865440,
    "text": "our problem of agents stepping on each"
  },
  {
    "startMs": 867480,
    "text": "other. Second one is we have a single"
  },
  {
    "startMs": 869960,
    "text": "source of truth"
  },
  {
    "startMs": 871440,
    "text": "with automatic conflict detection that"
  },
  {
    "startMs": 873720,
    "text": "is not element based but actually rule"
  },
  {
    "startMs": 875640,
    "text": "based that can detect that this agent"
  },
  {
    "startMs": 877800,
    "text": "did this issue."
  },
  {
    "startMs": 879120,
    "text": "And we can or want to change this value"
  },
  {
    "startMs": 881880,
    "text": "and actually resonate in the whole"
  },
  {
    "startMs": 883360,
    "text": "system immediately."
  },
  {
    "startMs": 884920,
    "text": "And thirdly, which I think of it as an"
  },
  {
    "startMs": 887080,
    "text": "IT administration for agent, we block at"
  },
  {
    "startMs": 889640,
    "text": "the source. Like we block from system"
  },
  {
    "startMs": 891600,
    "text": "level, not about level like tool by"
  },
  {
    "startMs": 893680,
    "text": "tool, but just we try to block it over"
  },
  {
    "startMs": 895680,
    "text": "there."
  },
  {
    "startMs": 897080,
    "text": "And the key lesson we learned here that"
  },
  {
    "startMs": 901480,
    "text": "agents care about like if you have your"
  },
  {
    "startMs": 904040,
    "text": "agents which are intelligent,"
  },
  {
    "startMs": 905880,
    "text": "it what matters is substrate layer that"
  },
  {
    "startMs": 907920,
    "text": "they are living in."
  },
  {
    "startMs": 909320,
    "text": "Like the world they living in is more"
  },
  {
    "startMs": 911360,
    "text": "important than the agents itself. Like"
  },
  {
    "startMs": 913400,
    "text": "what they can do, what they cannot do,"
  },
  {
    "startMs": 915360,
    "text": "what you allow and what you don't allow."
  },
  {
    "startMs": 919040,
    "text": "Cool. So, I'm going to use the word"
  },
  {
    "startMs": 921640,
    "text": "bottleneck. It's been used many times,"
  },
  {
    "startMs": 923800,
    "text": "but actually it's bottleneck in our"
  },
  {
    "startMs": 925120,
    "text": "case."
  },
  {
    "startMs": 926040,
    "text": "It wasn't missing intelligence. It was"
  },
  {
    "startMs": 927960,
    "text": "missing alignment."
  },
  {
    "startMs": 929480,
    "text": "And a shared nervous system lets your"
  },
  {
    "startMs": 931560,
    "text": "team move like a one body."
  },
  {
    "startMs": 933640,
    "text": "As we see at the moment, one of the"
  },
  {
    "startMs": 935200,
    "text": "things I like hearing from our subject"
  },
  {
    "startMs": 937600,
    "text": "matter experts that they're saying that"
  },
  {
    "startMs": 939720,
    "text": "at the beginning of system is not"
  },
  {
    "startMs": 941120,
    "text": "working fine."
  },
  {
    "startMs": 942760,
    "text": "Now it is good. Now I feel it's racing"
  },
  {
    "startMs": 945760,
    "text": "me. This is success for our case. And we"
  },
  {
    "startMs": 948800,
    "text": "think that this gives you four x"
  },
  {
    "startMs": 950440,
    "text": "leverage from our measurement at the"
  },
  {
    "startMs": 951840,
    "text": "moment."
  },
  {
    "startMs": 952760,
    "text": "And alignment is universal."
  },
  {
    "startMs": 955440,
    "text": "We're building it for the hardest case,"
  },
  {
    "startMs": 957640,
    "text": "which is ship design."
  },
  {
    "startMs": 960680,
    "text": "So currently we're in alpha stage"
  },
  {
    "startMs": 963640,
    "text": "with our development partners."
  },
  {
    "startMs": 966120,
    "text": "And the sign ups for beta are open. And"
  },
  {
    "startMs": 969440,
    "text": "you can actually join now and we expect"
  },
  {
    "startMs": 971880,
    "text": "it to release it in October 26."
  },
  {
    "startMs": 975560,
    "text": "If you want to"
  },
  {
    "startMs": 977240,
    "text": "reach out us, sign up for"
  },
  {
    "startMs": 980240,
    "text": "the beta,"
  },
  {
    "startMs": 981600,
    "text": "just use this QR code or the link over"
  },
  {
    "startMs": 983680,
    "text": "there. Thank you everyone."
  },
  {
    "startMs": 987751,
    "text": ">> [applause]"
  },
  {
    "startMs": 1003471,
    "text": "[music]"
  }
]
