[
  {
    "start": 0.18,
    "end": 4.14,
    "text": "As a software developer,\nand I guess in general (laughs) as a"
  },
  {
    "start": 4.22,
    "end": 8.2,
    "text": "human,\nbut especially as a software developer,"
  },
  {
    "start": 8.3,
    "end": 10.94,
    "text": "Anthropic right now, if you want to\nor not."
  },
  {
    "start": 10.95,
    "end": 14.89,
    "text": "And I don't think you should try to ignore\nit because it is"
  },
  {
    "start": 14.92,
    "end": 18.75,
    "text": "relevant. It's relevant for our future as"
  },
  {
    "start": 18.78,
    "end": 20.5,
    "text": "software developers, I would say."
  },
  {
    "start": 20.62,
    "end": 24.58,
    "text": "And in this episode,\nI won't talk about the Claude Code leak we"
  },
  {
    "start": 24.64,
    "end": 25.66,
    "text": "had last week."
  },
  {
    "start": 26.6,
    "end": 29.8,
    "text": "I won't talk about their, uh,"
  },
  {
    "start": 29.94,
    "end": 33.9,
    "text": "reinforced terms regarding the usage of\ntheir subscription"
  },
  {
    "start": 33.96,
    "end": 37.65,
    "text": "offerings, Claude Max and so on,\nand how they're"
  },
  {
    "start": 37.74,
    "end": 41.56,
    "text": "cracking down on unauthorized use of those"
  },
  {
    "start": 41.6,
    "end": 45.58,
    "text": "subscriptions. They are doing\nthat right now because, of course, their"
  },
  {
    "start": 45.62,
    "end": 49.22,
    "text": "subscription offerings,\njust like the ones by OpenAI, are heavily"
  },
  {
    "start": 49.42,
    "end": 53.34,
    "text": "subsidized and they can't make any money\nif everybody"
  },
  {
    "start": 53.48,
    "end": 57.0,
    "text": "maxes out their subscriptions. So yeah,\nthey're really"
  },
  {
    "start": 57.1,
    "end": 60.74,
    "text": "restricting,\nor they're trying to restrict,"
  },
  {
    "start": 60.75,
    "end": 64.58,
    "text": "subscriptions to just humans,\njust on their"
  },
  {
    "start": 64.64,
    "end": 68.55,
    "text": "website, or in Claude Code\nor the Claude desktop app, I"
  },
  {
    "start": 68.64,
    "end": 72.53,
    "text": "guess. Uh, but yeah, again, this\nis not the focus here and I"
  },
  {
    "start": 72.56,
    "end": 76.2,
    "text": "won't even focus on their impressive\nrevenue"
  },
  {
    "start": 76.32,
    "end": 79.95,
    "text": "growth, which is worth, uh, a,\na short note though, because"
  },
  {
    "start": 80.02,
    "end": 83.66,
    "text": "Anthropic has reached a annual recurring"
  },
  {
    "start": 83.88,
    "end": 87.53,
    "text": "revenue of $30 billion, which is"
  },
  {
    "start": 87.58,
    "end": 91.24,
    "text": "already impressive,\nbut especially impressive"
  },
  {
    "start": 91.68,
    "end": 95.32,
    "text": "$9 billion at the end of 2025. So, they"
  },
  {
    "start": 95.86,
    "end": 99.64,
    "text": "more than tripled their annual recurring\nrevenue within"
  },
  {
    "start": 99.74,
    "end": 102.58,
    "text": "just a few months, which\nis really impressive."
  },
  {
    "start": 102.64,
    "end": 106.15,
    "text": "And therefore, of course,\nif you wanna learn how to efficiently"
  },
  {
    "start": 106.64,
    "end": 109.96,
    "text": "use Claude Code,\nhow to get the most out of it,"
  },
  {
    "start": 110.37,
    "end": 114.21,
    "text": "And, um, it's also very popular,\nwhich of course makes me happy,"
  },
  {
    "start": 114.28,
    "end": 118.1,
    "text": "and you'll find a link below\nif you wanna join it and learn how to"
  },
  {
    "start": 118.14,
    "end": 120.3,
    "text": "efficiently work with Claude Code."
  },
  {
    "start": 120.36,
    "end": 123.8,
    "text": "But as mentioned,\nthat's not even the main topic here."
  },
  {
    "start": 123.88,
    "end": 127.7,
    "text": "Instead,\nI wanna talk about Project Glasswing and"
  },
  {
    "start": 127.74,
    "end": 131.48,
    "text": "their new model, Mythos,\nwhich they haven't"
  },
  {
    "start": 131.56,
    "end": 135.32,
    "text": "released to the public,\nand they also shared"
  },
  {
    "start": 135.4,
    "end": 139.02,
    "text": "why. And I think this\nis important to understand, and it's also"
  },
  {
    "start": 139.04,
    "end": 142.9,
    "text": "important to try to look behind the\nscenes, behind their"
  },
  {
    "start": 142.92,
    "end": 146.78,
    "text": "rationale,\nand what the impact of this new model"
  },
  {
    "start": 146.88,
    "end": 150.46,
    "text": "and how it works and its capabilities\nis for us developers."
  },
  {
    "start": 150.53,
    "end": 153.89,
    "text": "So, what is Project Glasswing? What\nis their new model all about?"
  },
  {
    "start": 154.58,
    "end": 158.04,
    "text": "Uh, below, of course,\nyou'll find a link to this article as"
  },
  {
    "start": 158.14,
    "end": 161.88,
    "text": "article on the official Anthropic site\nwhere they announce Project Glasswing"
  },
  {
    "start": 162.0,
    "end": 164.8,
    "text": "and also talk about their new model."
  },
  {
    "start": 164.84,
    "end": 168.4,
    "text": "And if I scroll down a bit,\nwe can already see some, uh,"
  },
  {
    "start": 168.6,
    "end": 172.5,
    "text": "summary benchmark statistics here,\nwhere we can see"
  },
  {
    "start": 172.54,
    "end": 176.47,
    "text": "that this new model, the Mythos, uh,\npreview version of the model, the,"
  },
  {
    "start": 176.58,
    "end": 180.26,
    "text": "the, the model name is Mythos, um,\nperforms way better than"
  },
  {
    "start": 180.34,
    "end": 184.32,
    "text": "Opus 4.6. And depending on which"
  },
  {
    "start": 184.38,
    "end": 188.32,
    "text": "exact benchmark you look on, there\nis quite a big difference"
  },
  {
    "start": 188.36,
    "end": 191.66,
    "text": "between Opus 4.6 and this new model."
  },
  {
    "start": 192.6,
    "end": 196.13,
    "text": "Now, of course, this is, uh, on its own,\nnot super impressive."
  },
  {
    "start": 196.22,
    "end": 200.14,
    "text": "Whenever a new model is announced,\nno matter by which company, it"
  },
  {
    "start": 200.22,
    "end": 204.1,
    "text": "performs way better,\nor at least a tiny bit better,"
  },
  {
    "start": 204.16,
    "end": 206.18,
    "text": "models,\notherwise it wouldn't get released."
  },
  {
    "start": 206.2,
    "end": 209.86,
    "text": "And of course, there\nare ways of gaming some of these"
  },
  {
    "start": 209.94,
    "end": 213.6,
    "text": "so I typically don't care too much about\nthose benchmark numbers,"
  },
  {
    "start": 213.66,
    "end": 217.52,
    "text": "and that wouldn't really be different for\nthis"
  },
  {
    "start": 217.56,
    "end": 221.4,
    "text": "model here, but there\nare interesting things about the"
  },
  {
    "start": 221.44,
    "end": 225.18,
    "text": "new Mythos model. And that's the fact\nthat Anthropic"
  },
  {
    "start": 225.22,
    "end": 229.1,
    "text": "decided to not release it to the public\nbecause, as"
  },
  {
    "start": 229.18,
    "end": 229.88,
    "text": "they say,"
  },
  {
    "start": 231.06,
    "end": 234.28,
    "text": "it is too good at finding and"
  },
  {
    "start": 234.36,
    "end": 238.34,
    "text": "exploiting vulnerabilities in operating"
  },
  {
    "start": 238.42,
    "end": 242.32,
    "text": "systems, any other software, browsers,\nreally, in"
  },
  {
    "start": 242.68,
    "end": 246.22,
    "text": "software in general. And in this article,\nand also in a"
  },
  {
    "start": 246.4,
    "end": 248.44,
    "text": "separate article, which\nis also linked below,"
  },
  {
    "start": 249.24,
    "end": 252.68,
    "text": "they share some details,\nand especially this separate article is"
  },
  {
    "start": 252.74,
    "end": 256.399,
    "text": "extremely long and gives concrete"
  },
  {
    "start": 256.459,
    "end": 259.74,
    "text": "examples for vulnerabilities and potential"
  },
  {
    "start": 259.76,
    "end": 263.42,
    "text": "exploits this new model found."
  },
  {
    "start": 264.2,
    "end": 267.74,
    "text": "Uh, for example,\nthey start in this article with a"
  },
  {
    "start": 267.78,
    "end": 271.05,
    "text": "very serious exploit and vulnerability"
  },
  {
    "start": 271.62,
    "end": 274.56,
    "text": "that was found in, uh,"
  },
  {
    "start": 274.64,
    "end": 278.6,
    "text": "OpenBSD. OpenBSD is, of course,\nan operating system which is"
  },
  {
    "start": 278.64,
    "end": 282.04,
    "text": "popular on certain pieces of networking\nsoftware, for example,"
  },
  {
    "start": 282.54,
    "end": 285.98,
    "text": "and Mythos, their new model running in"
  },
  {
    "start": 286.06,
    "end": 289.86,
    "text": "Agentic Harness, like Claude Code,\nI guess, was able"
  },
  {
    "start": 289.94,
    "end": 293.6,
    "text": "to find and exploit, a- and that's the"
  },
  {
    "start": 293.62,
    "end": 297.17,
    "text": "interesting part - a vulnerability\nthat was related"
  },
  {
    "start": 297.22,
    "end": 300.86,
    "text": "to integer overflow and, uh, memory"
  },
  {
    "start": 300.94,
    "end": 304.68,
    "text": "access, unexpected memory access, um,"
  },
  {
    "start": 305.04,
    "end": 308.4,
    "text": "that was able to crash machines that\nwere running"
  },
  {
    "start": 308.46,
    "end": 312.4,
    "text": "OpenBSD in a reproducible way, which"
  },
  {
    "start": 312.41,
    "end": 316.4,
    "text": "could, of course,\nbe leveraged to run very damaging,"
  },
  {
    "start": 316.52,
    "end": 319.94,
    "text": "uh, denial of service attacks by over\nand over"
  },
  {
    "start": 320.0,
    "end": 323.8,
    "text": "sending specific packets\nand requests to such machines"
  },
  {
    "start": 323.82,
    "end": 327.4,
    "text": "here, which exploited\nthat vulnerability to bring"
  },
  {
    "start": 327.44,
    "end": 331.28,
    "text": "down those machines\nand potentially bring down entire"
  },
  {
    "start": 331.32,
    "end": 334.98,
    "text": "corporate networks.\nAnd this vulnerability was"
  },
  {
    "start": 335.0,
    "end": 338.72,
    "text": "detected in a run that cost under $50,\nthough"
  },
  {
    "start": 339.34,
    "end": 342.6,
    "text": "the overall set of runs cost under"
  },
  {
    "start": 342.62,
    "end": 346.36,
    "text": "$20,000. And since you, of course,\ndon't know in advance which"
  },
  {
    "start": 346.5,
    "end": 350.16,
    "text": "run will find a vulnerability, it's\nthat number that matters."
  },
  {
    "start": 351.24,
    "end": 355.11,
    "text": "Still, of course,\nit's easy to imagine that"
  },
  {
    "start": 355.18,
    "end": 358.84,
    "text": "a model capable of finding such critical"
  },
  {
    "start": 358.88,
    "end": 362.28,
    "text": "vulnerabilities for such a comparatively\nlow"
  },
  {
    "start": 362.36,
    "end": 366.31,
    "text": "cost...Depending on who you are,\nif you're a nation, um, for"
  },
  {
    "start": 366.38,
    "end": 368.27,
    "text": "example, or some, uh,"
  },
  {
    "start": 369.12,
    "end": 373.0,
    "text": "serious bad actor,\nthat may not be a lot of money"
  },
  {
    "start": 373.06,
    "end": 373.4,
    "text": "to you."
  },
  {
    "start": 374.21,
    "end": 377.54,
    "text": "That, of course,\nis a problem because it's easy to imagine"
  },
  {
    "start": 377.58,
    "end": 381.04,
    "text": "that if such a model were"
  },
  {
    "start": 381.06,
    "end": 384.74,
    "text": "developed by a company, an organization"
  },
  {
    "start": 385.66,
    "end": 388.74,
    "text": "that cares a bit less about security"
  },
  {
    "start": 389.2,
    "end": 392.51,
    "text": "and/or that maybe doesn't have to"
  },
  {
    "start": 392.64,
    "end": 395.9,
    "text": "fear any consequences of"
  },
  {
    "start": 395.98,
    "end": 399.06,
    "text": "abusing such vulnerabilities,"
  },
  {
    "start": 400.16,
    "end": 403.88,
    "text": "that this could be a problem. And it seems"
  },
  {
    "start": 404.34,
    "end": 408.24,
    "text": "as if we're entering a new age"
  },
  {
    "start": 408.64,
    "end": 411.2,
    "text": "with AI, with these AI models,"
  },
  {
    "start": 412.92,
    "end": 416.76,
    "text": "where nothing is secure and"
  },
  {
    "start": 416.82,
    "end": 420.18,
    "text": "it's easier than ever to mass"
  },
  {
    "start": 420.2,
    "end": 423.94,
    "text": "deploy AI agents running models like this\nto"
  },
  {
    "start": 424.04,
    "end": 427.63,
    "text": "scan all kinds of pieces of software\nand find and"
  },
  {
    "start": 427.64,
    "end": 429.64,
    "text": "potentially exploit vulnerabilities."
  },
  {
    "start": 429.68,
    "end": 433.39,
    "text": "And of course, as a human on, on your own,"
  },
  {
    "start": 433.54,
    "end": 435.96,
    "text": "there is no way of keeping up with it."
  },
  {
    "start": 436.04,
    "end": 439.66,
    "text": "I mean, the bug, the exploit that\nwas found here existed"
  },
  {
    "start": 439.74,
    "end": 443.7,
    "text": "for, I think they said 27 years\nor something like this."
  },
  {
    "start": 445.04,
    "end": 448.62,
    "text": "This shows that no human was able to"
  },
  {
    "start": 448.7,
    "end": 452.58,
    "text": "find this bug in such a long period of\ntime, including"
  },
  {
    "start": 452.62,
    "end": 456.38,
    "text": "bad actors,\nwhich of course would've had an interest"
  },
  {
    "start": 456.39,
    "end": 459.4,
    "text": "attack this operating system in the past\ntoo."
  },
  {
    "start": 459.8,
    "end": 463.5,
    "text": "Now this is just one,\nmaybe the most prominent"
  },
  {
    "start": 463.58,
    "end": 467.38,
    "text": "finding, um, this new model had. They\nare listing"
  },
  {
    "start": 467.6,
    "end": 471.5,
    "text": "way more bugs and exploits the model\nwas find and"
  },
  {
    "start": 471.58,
    "end": 473.46,
    "text": "sometimes also able to abuse,"
  },
  {
    "start": 474.5,
    "end": 476.32,
    "text": "and they also shared other stories."
  },
  {
    "start": 476.6,
    "end": 479.92,
    "text": "Uh, on X, for example,\nlike the model being able to"
  },
  {
    "start": 479.98,
    "end": 483.95,
    "text": "escape a sandbox or the AI agent running\nthe"
  },
  {
    "start": 484.04,
    "end": 487.42,
    "text": "model being able to escape a sandbox in\nwhich it was running."
  },
  {
    "start": 488.3,
    "end": 491.979,
    "text": "And that brings us back to Project\nGlasswing, which is an"
  },
  {
    "start": 492.04,
    "end": 495.98,
    "text": "initiative created by Anthropic together"
  },
  {
    "start": 496.14,
    "end": 499.26,
    "text": "with other big companies like AWS, Apple,"
  },
  {
    "start": 499.92,
    "end": 503.48,
    "text": "um, Microsoft, the Linux Foundation,\nand others"
  },
  {
    "start": 503.62,
    "end": 507.2,
    "text": "to use this model to basically patch"
  },
  {
    "start": 507.3,
    "end": 511.19,
    "text": "up their software before this is"
  },
  {
    "start": 511.22,
    "end": 514.979,
    "text": "publicly released\nand before the public gets their hands"
  },
  {
    "start": 515.05,
    "end": 518.9,
    "text": "on this model. That is the,\nthe narrative of this article."
  },
  {
    "start": 519.02,
    "end": 521.32,
    "text": "That is the explanation of Anthropic."
  },
  {
    "start": 522.84,
    "end": 523.229,
    "text": "And"
  },
  {
    "start": 524.14,
    "end": 527.1,
    "text": "I have some mixed thoughts here. Now,\nfor one,"
  },
  {
    "start": 528.32,
    "end": 532.04,
    "text": "I have no strong reason to believe\nthat this isn't true."
  },
  {
    "start": 532.42,
    "end": 536.0,
    "text": "Clearly, Anthropic would have some"
  },
  {
    "start": 536.18,
    "end": 540.08,
    "text": "reasons to not release this model outside\nof what they're"
  },
  {
    "start": 540.14,
    "end": 541.52,
    "text": "mentioning here. For example,"
  },
  {
    "start": 542.34,
    "end": 546.02,
    "text": "um, I read that this model is a roughly"
  },
  {
    "start": 546.16,
    "end": 550.14,
    "text": "10 trillion parameters model, which\nis way bigger than"
  },
  {
    "start": 550.24,
    "end": 554.12,
    "text": "all the frontier models we had thus far\nor were able"
  },
  {
    "start": 554.16,
    "end": 556.08,
    "text": "to use publicly thus far,"
  },
  {
    "start": 556.96,
    "end": 560.64,
    "text": "and training it is said to have cost\naround"
  },
  {
    "start": 560.66,
    "end": 564.56,
    "text": "$10 billion. The token cost of"
  },
  {
    "start": 564.62,
    "end": 568.38,
    "text": "this model, I read, is expected to"
  },
  {
    "start": 568.42,
    "end": 571.38,
    "text": "be around this range, $25,"
  },
  {
    "start": 571.43,
    "end": 575.3,
    "text": "$125 for, uh, input and output tokens."
  },
  {
    "start": 575.58,
    "end": 579.32,
    "text": "And of course,\nthat would also be reasons for not"
  },
  {
    "start": 579.36,
    "end": 583.22,
    "text": "releasing this model because they can't\ninclude it in"
  },
  {
    "start": 583.24,
    "end": 586.67,
    "text": "their Claude subscriptions because it's\njust too"
  },
  {
    "start": 586.7,
    "end": 590.62,
    "text": "expensive.\nThey would have to ramp up the"
  },
  {
    "start": 590.66,
    "end": 593.55,
    "text": "to a price point where not a lot of people\nare willing to pay it,"
  },
  {
    "start": 594.81,
    "end": 598.42,
    "text": "uh,\nand therefore there wouldn't really be a"
  },
  {
    "start": 598.46,
    "end": 601.64,
    "text": "exposing it to the public,\nat least as part of Claude code."
  },
  {
    "start": 601.68,
    "end": 605.449,
    "text": "Now of course,\nthey could still expose it through their"
  },
  {
    "start": 605.48,
    "end": 609.4,
    "text": "pay-per-use cost basis, and\nif it's expensive, who"
  },
  {
    "start": 609.46,
    "end": 613.24,
    "text": "cares? If there are companies, people\nthat would be willing to pay it,"
  },
  {
    "start": 613.46,
    "end": 616.74,
    "text": "they could do that. And that, of course,\nis the part where the"
  },
  {
    "start": 616.78,
    "end": 620.02,
    "text": "cybersecurity concerns might really come\ninto play"
  },
  {
    "start": 620.079,
    "end": 623.88,
    "text": "because clearly that all"
  },
  {
    "start": 623.94,
    "end": 627.4,
    "text": "is very likely not made up. I mean,\nit's definitely not made"
  },
  {
    "start": 627.48,
    "end": 631.22,
    "text": "up. The FFmpeg team, for example, which"
  },
  {
    "start": 631.3,
    "end": 635.16,
    "text": "is also listed here as a, um, as a"
  },
  {
    "start": 635.22,
    "end": 638.9,
    "text": "vulnerability that they were f-\nable to find a vulnerability in"
  },
  {
    "start": 638.98,
    "end": 639.74,
    "text": "FFmpeg,"
  },
  {
    "start": 640.6,
    "end": 644.0,
    "text": "uh, the team confirmed on X that Anthropic"
  },
  {
    "start": 644.18,
    "end": 648.0,
    "text": "sent, uh, sent a patch for, uh, uh, a"
  },
  {
    "start": 648.04,
    "end": 651.05,
    "text": "vulnerability in the FFmpeg, uh,"
  },
  {
    "start": 651.3,
    "end": 654.96,
    "text": "software, uh, program So yeah, this is"
  },
  {
    "start": 654.98,
    "end": 658.6,
    "text": "clearly not made up. These concerns\nare valid."
  },
  {
    "start": 659.08,
    "end": 662.5,
    "text": "Cybersecurity concerns are valid,\nespecially since, of course,"
  },
  {
    "start": 663.06,
    "end": 666.78,
    "text": "if money is not the main issue,\nyou could deploy thousands of"
  },
  {
    "start": 666.98,
    "end": 670.82,
    "text": "agents running simultaneously using this\nor"
  },
  {
    "start": 670.94,
    "end": 674.66,
    "text": "similar models,\nwhich we may have in the future to, yeah,"
  },
  {
    "start": 675.0,
    "end": 678.52,
    "text": "scan all kinds of software\nand exploit them."
  },
  {
    "start": 678.56,
    "end": 681.0,
    "text": "And of course, the big problem is that"
  },
  {
    "start": 682.56,
    "end": 686.4,
    "text": "using this model to find vulnerabilities"
  },
  {
    "start": 686.44,
    "end": 689.35,
    "text": "and patch them up is possible,"
  },
  {
    "start": 690.18,
    "end": 693.78,
    "text": "but it's only possible if the"
  },
  {
    "start": 693.88,
    "end": 697.6,
    "text": "owner or the maintainer of a certain piece\nof software can afford the"
  },
  {
    "start": 697.66,
    "end": 700.9,
    "text": "model or gets access for free\nor anything like that."
  },
  {
    "start": 701.24,
    "end": 705.02,
    "text": "And even if a vulnerability is patched up,"
  },
  {
    "start": 705.18,
    "end": 705.9,
    "text": "we all know"
  },
  {
    "start": 706.86,
    "end": 710.76,
    "text": "that not all (laughs) computers out there,\nnot all machines,"
  },
  {
    "start": 710.86,
    "end": 712.08,
    "text": "not all users"
  },
  {
    "start": 713.14,
    "end": 717.0,
    "text": "have up-to-date software running on them."
  },
  {
    "start": 717.12,
    "end": 720.26,
    "text": "If you were to take a look at all the"
  },
  {
    "start": 720.28,
    "end": 723.64,
    "text": "various servers running out there in the"
  },
  {
    "start": 723.68,
    "end": 727.38,
    "text": "worldwide webI would guess the vast"
  },
  {
    "start": 727.42,
    "end": 730.66,
    "text": "majority of them is running outdated\nsoftware."
  },
  {
    "start": 730.76,
    "end": 734.46,
    "text": "I mean, on our phones, our, our laptops,\nwe are"
  },
  {
    "start": 734.9,
    "end": 738.46,
    "text": "of- not running the latest software,\nthe latest"
  },
  {
    "start": 738.86,
    "end": 742.76,
    "text": "version of our operating system.\nThe latest security patch may not be"
  },
  {
    "start": 742.78,
    "end": 746.44,
    "text": "installed and that\nis true for all layers of s- of"
  },
  {
    "start": 746.52,
    "end": 750.46,
    "text": "software.\nAnd in a world where it's easier than ever"
  },
  {
    "start": 750.58,
    "end": 754.54,
    "text": "find security vulnerabilities,\nthat of course becomes an"
  },
  {
    "start": 754.64,
    "end": 758.39,
    "text": "even bigger issue.\nBecause of course the good thing about"
  },
  {
    "start": 758.46,
    "end": 761.88,
    "text": "this AI model is that it can also be"
  },
  {
    "start": 761.96,
    "end": 765.8,
    "text": "used for proactively looking for security"
  },
  {
    "start": 765.9,
    "end": 769.8,
    "text": "vulnerabilities and patching them.\nSo it's not just a tool for attackers."
  },
  {
    "start": 769.84,
    "end": 773.62,
    "text": "It can also make defense easier because\nyou now have a"
  },
  {
    "start": 773.68,
    "end": 777.52,
    "text": "tool that can be run simultaneously in\nparallel across"
  },
  {
    "start": 777.53,
    "end": 780.38,
    "text": "thousands of agents to make your software\nsecure."
  },
  {
    "start": 780.44,
    "end": 784.38,
    "text": "In theory,\nthis can be a very useful tool for"
  },
  {
    "start": 784.48,
    "end": 788.34,
    "text": "of course, again, not every company,\nperson"
  },
  {
    "start": 788.4,
    "end": 792.26,
    "text": "that may be developing crucial software\nmay be able to afford it, may"
  },
  {
    "start": 792.34,
    "end": 796.24,
    "text": "be interested in using it,\nand then even if it is"
  },
  {
    "start": 796.32,
    "end": 798.9,
    "text": "used to find and patch vulnerabilities,"
  },
  {
    "start": 799.54,
    "end": 803.34,
    "text": "still these latest versions will not be\ninstalled"
  },
  {
    "start": 803.4,
    "end": 807.22,
    "text": "everywhere.\nAnd that of course gives attackers, uh,"
  },
  {
    "start": 807.3,
    "end": 811.03,
    "text": "opportunity where they know about way"
  },
  {
    "start": 811.12,
    "end": 814.68,
    "text": "more vulnerabilities than before at some\npoint because"
  },
  {
    "start": 815.56,
    "end": 819.49,
    "text": "way more vulnerabilities are detected,\nbut not every"
  },
  {
    "start": 819.52,
    "end": 822.96,
    "text": "machine, every user\nis protected against those"
  },
  {
    "start": 822.98,
    "end": 826.73,
    "text": "vulnerabilities and that\nis one real concern I have"
  },
  {
    "start": 826.82,
    "end": 830.6,
    "text": "about this development. Now that is the,\nthe bigger"
  },
  {
    "start": 830.64,
    "end": 834.48,
    "text": "picture which affects everybody,\nall companies, all humans"
  },
  {
    "start": 834.56,
    "end": 838.1,
    "text": "in the end. Another question of course\nis what does"
  },
  {
    "start": 838.58,
    "end": 842.0,
    "text": "a model like this mean for us"
  },
  {
    "start": 842.32,
    "end": 845.76,
    "text": "developers? I mean,\nclearly this seems to be a"
  },
  {
    "start": 845.94,
    "end": 849.82,
    "text": "highly capable model that\nwas able to look for"
  },
  {
    "start": 850.04,
    "end": 854.0,
    "text": "vulnerabilities on its own\nand exploit vulnerabilities on its"
  },
  {
    "start": 854.06,
    "end": 857.9,
    "text": "own. So yeah, what is the impact for"
  },
  {
    "start": 857.94,
    "end": 858.7,
    "text": "developers?"
  },
  {
    "start": 859.82,
    "end": 862.88,
    "text": "And I think here, when it comes to that,"
  },
  {
    "start": 864.18,
    "end": 868.02,
    "text": "not a lot changes for now. I"
  },
  {
    "start": 868.08,
    "end": 871.84,
    "text": "mean,\nwe're already living in a world where"
  },
  {
    "start": 871.92,
    "end": 875.6,
    "text": "AI agents like Claude code\nand the underlying models,"
  },
  {
    "start": 875.64,
    "end": 879.53,
    "text": "true for Codex and so on,\nwhatever your favorite AI agent"
  },
  {
    "start": 880.28,
    "end": 882.8,
    "text": "are able to generate most of our code."
  },
  {
    "start": 883.2,
    "end": 885.28,
    "text": "You may not be using them,\nyou may not like them."
  },
  {
    "start": 885.6,
    "end": 889.51,
    "text": "I created a separate video where I share\nmy feelings about that and that"
  },
  {
    "start": 889.58,
    "end": 893.56,
    "text": "this sucks the joy out of the,\nthe software development part"
  },
  {
    "start": 893.6,
    "end": 897.37,
    "text": "for me, but it is the,\nthe reality nonetheless, if you like it or"
  },
  {
    "start": 897.46,
    "end": 901.28,
    "text": "not. And believe me,\nI don't necessarily like it, but yeah,"
  },
  {
    "start": 901.34,
    "end": 905.28,
    "text": "reality. Nonetheless,\nwhat a human brings to the"
  },
  {
    "start": 905.32,
    "end": 909.22,
    "text": "table or why humans still matter here\nand may matter"
  },
  {
    "start": 909.42,
    "end": 913.32,
    "text": "more than ever is, of course,\nthat you definitely"
  },
  {
    "start": 913.329,
    "end": 917.2,
    "text": "don't want an AI agent like this go rogue"
  },
  {
    "start": 917.26,
    "end": 920.58,
    "text": "and work totally on its own. Steering"
  },
  {
    "start": 921.0,
    "end": 924.56,
    "text": "such models and agents, controlling them,"
  },
  {
    "start": 924.78,
    "end": 928.54,
    "text": "giving them clear tasks,\nlimiting the scope of the"
  },
  {
    "start": 928.62,
    "end": 929.54,
    "text": "work they do,"
  },
  {
    "start": 931.0,
    "end": 934.34,
    "text": "all these things are more important than\never."
  },
  {
    "start": 934.549,
    "end": 938.22,
    "text": "These models can, as it seems, do way"
  },
  {
    "start": 938.32,
    "end": 939.14,
    "text": "more than"
  },
  {
    "start": 939.98,
    "end": 943.82,
    "text": "the vast majority of developers can do.\nDefinitely way more than I can do."
  },
  {
    "start": 945.42,
    "end": 949.16,
    "text": "And yet,\nwhen it comes to shipping products,"
  },
  {
    "start": 949.26,
    "end": 952.7,
    "text": "comes to building software used by humans,"
  },
  {
    "start": 954.36,
    "end": 957.96,
    "text": "the influence of a human is of"
  },
  {
    "start": 958.06,
    "end": 960.5,
    "text": "course extremely, extremely important."
  },
  {
    "start": 960.6,
    "end": 964.3,
    "text": "What's changing, of course,\nis our role as software developers."
  },
  {
    "start": 964.4,
    "end": 968.26,
    "text": "We're changing from the people that\nare writing the code to the"
  },
  {
    "start": 968.32,
    "end": 971.62,
    "text": "people that are steering the model,\nthat are reviewing the code, that are"
  },
  {
    "start": 971.72,
    "end": 975.34,
    "text": "understanding what it does, that\nare setting the scope."
  },
  {
    "start": 975.4,
    "end": 979.34,
    "text": "And yeah, again,\nI talked about this in that other video,"
  },
  {
    "start": 979.4,
    "end": 982.94,
    "text": "and then that this may not necessarily be\nwhat, uh,"
  },
  {
    "start": 983.179,
    "end": 986.92,
    "text": "what, what you like.\nIt's definitely not w- why I got"
  },
  {
    "start": 986.94,
    "end": 988.8,
    "text": "into software development in the first\nplace."
  },
  {
    "start": 988.88,
    "end": 992.62,
    "text": "But yeah, this is the impact here\nand the more capable these"
  },
  {
    "start": 992.66,
    "end": 996.6,
    "text": "models get,\nthe more important I think it gets to"
  },
  {
    "start": 996.78,
    "end": 1000.51,
    "text": "have that,\nthat human voice in there as well,"
  },
  {
    "start": 1000.6,
    "end": 1004.2,
    "text": "that human influence, uh, in there as well"
  },
  {
    "start": 1005.28,
    "end": 1008.62,
    "text": "So that is, that is\nthat changing role and, and"
  },
  {
    "start": 1009.32,
    "end": 1013.02,
    "text": "our role in the future. But yeah, I mean,"
  },
  {
    "start": 1013.9,
    "end": 1017.16,
    "text": "these are really interesting developments\nand"
  },
  {
    "start": 1017.78,
    "end": 1021.68,
    "text": "especially this model\nand its implications and"
  },
  {
    "start": 1021.86,
    "end": 1025.13,
    "text": "that cybersecurity relevance it has"
  },
  {
    "start": 1026.08,
    "end": 1029.36,
    "text": "makes one think what would have happened\nor what would"
  },
  {
    "start": 1029.42,
    "end": 1032.49,
    "text": "happen if other"
  },
  {
    "start": 1032.96,
    "end": 1036.11,
    "text": "actors, other nations or"
  },
  {
    "start": 1036.18,
    "end": 1039.92,
    "text": "organizations in the world get their hands"
  },
  {
    "start": 1040.02,
    "end": 1043.76,
    "text": "on this model or models that\nare similar in capability."
  },
  {
    "start": 1043.8,
    "end": 1047.06,
    "text": "Because of course it's only a matter of"
  },
  {
    "start": 1047.14,
    "end": 1050.63,
    "text": "time until models with similar"
  },
  {
    "start": 1050.66,
    "end": 1053.36,
    "text": "capabilities are accessible"
  },
  {
    "start": 1054.24,
    "end": 1057.9,
    "text": "by the public or certainly at least by"
  },
  {
    "start": 1058.84,
    "end": 1061.44,
    "text": "other nations and actors. And"
  },
  {
    "start": 1062.38,
    "end": 1066.12,
    "text": "yeah, I'm not sure if we\nare prepared for that"
  },
  {
    "start": 1066.24,
    "end": 1070.22,
    "text": "new race in cybersecurity and that delay"
  },
  {
    "start": 1070.26,
    "end": 1074.24,
    "text": "between bugs being found and being patched"
  },
  {
    "start": 1074.42,
    "end": 1077.84,
    "text": "and people installing those patches."
  },
  {
    "start": 1078.16,
    "end": 1082.14,
    "text": "I think we'll enter a new era of\ncybersecurity and we'll"
  },
  {
    "start": 1082.2,
    "end": 1086.18,
    "text": "be able to adjust, I'm sure, but this, uh,\ndefinitely"
  },
  {
    "start": 1086.24,
    "end": 1087.71,
    "text": "marks an interesting,"
  },
  {
    "start": 1088.6,
    "end": 1092.51,
    "text": "uh, point in the history of,\nof model development"
  },
  {
    "start": 1093.24,
    "end": 1093.64,
    "text": "I would say"
  }
]