{
  "text": "You've probably typed something into\nGemini, got an answer, and closed the\ntab, the same way you'd use any other\nchatbot. Here's the thing, that's maybe\n10% of what it actually does. Use it\nright, and it can quietly take over\nalmost half the busy work you're still\ndoing by hand. I spent hours mapping\nevery model, every mode, and every\nproduct Gemini is quietly wired into,\nand the number that stopped me was this.\nGemini's app alone has over 650 million\nmonthly users, and that's before you\ncount everyone using it inside search.\nMost of them are using maybe 20% of it,\nwith no idea the rest even exists. Look\nat this data from the Census Bureau.\nOnly about one in five US businesses\nactually use AI in their operations.\nSo, if you run a business, and you're\neven thinking about this, you're ahead\nof most of your competition.\nWhat you might not know is that\nalongside covering AI news, we work with\nbusiness owners to help them implement\nAI in their business.\nOur engineering team gets to know how\nyour business runs, then builds the\nautomation with you.\nYou'll find the link in the description\nbelow.\nbelow.\nClick it, fill out a short form about\nyour business, and we'll get in touch to\nset up a call. So, in this video, I'm\nbreaking down exactly what Gemini is as\nof mid-2026,\nevery current model, every mode, and\neverywhere Google has quietly built it\nin. By the end, you'll know exactly\nwhich Gemini tool to reach for depending\non what you're actually trying to do,\ninstead of just typing into whichever\nbox is in front of you. First, let's\nclear up the biggest misconception.\nGemini isn't one product at all. What\nGemini actually is, here's the mental\nmodel you need before any of this makes\nsense. Gemini isn't a single AI, it's\nGoogle's umbrella name for a whole\nplatform, a family of models underneath,\nand a set of products on top that let\nyou actually talk to them. Think of it\nin two layers. The bottom layer is the\nmodels themselves, things like Gemini\n3.6 Flash or Gemini 3.1 Pro. These are\nthe engines tuned for different jobs.\nSome built for speed, some for heavy\nreasoning, some for images or audio. You\nnever see these names unless you go\nlooking. The top layer is everything you\nactually click on. The Gemini app, AI\nmode inside Google search, Gemini inside\nGmail and Docs, the voice assistant on\nyour phone. All of those are just\ndifferent doors into the same underlying\nmodels. That's the whole point of this\nvideo. Google isn't trying to build one\ngreat chatbot. It's trying to put the\nsame AI brain behind every product you\nalready use. So, let's start with the\nbrains, the actual models, because once\nyou know what each one is built for,\neverything else clicks into place. The\ncurrent model lineup. This is a demo\nchecklist, so we're going model by\nmodel. What it is, what it's actually\ngood for, and where you can get it.\nGemini 3.7 flash.\nLaunched on August 13th, 2026, this is\nGoogle's newest flash model and its most\ncapable workhorse yet. It's built\nprimarily for coding and AI agents with\nmajor improvements in software\nengineering, web development, and\ncomplex multi-step workflows. Google has\nalready made it generally available\nthrough the Gemini API, positioning 3.7\nflash as the new go-to model when you\nwant strong intelligence without giving\nup the speed and efficiency the flash\nlineup is known for. Gemini 3.6 flash.\nThis is Google's current flagship,\nannounced in a company blog post on July\n21st, 2026. It's built as a workhorse,\nstrong at coding, knowledge work, and\nmultimodal tasks. And according to\nGoogle's own numbers, it does the job\nusing about 17% fewer tokens on average\nthan its predecessor. Fewer tokens means\nfaster answers and a lower bill if\nyou're paying for it through the API.\nYou can reach it through the Gemini API,\nthrough AI Studio, or simply by using\nthe Gemini app and searches AI mode. No\nextra setup required. If you only\nremember one model name from this video,\nmake it this one because it's what most\nof Gemini is quietly running on right\nnow. Gemini 3.5 Flash. This one launched\nback in May 2026\nand it's the model that was actually\npowering AI mode in search before 3.6\nFlash took over. Google described it as\nfrontier-level intelligence at\nexceptional speed and by its own\nbenchmarks, it pushed output throughput\nto roughly four times faster than other\ntop models at the time. It's still very\nmuch in active use across the Gemini\napp, Google's anti-gravity platform, and\nenterprise tools. It's slightly behind\n3.6 now, but it's the model quietly\nsitting behind a huge share of what\nshipped this year. Gemini 3.5\nFlashlight. Same July announcement,\ndifferent job entirely. This is the\nstripped-down, high-throughput sibling.\nGoogle sites roughly 350 tokens per\nsecond, which is built for volume, not\ndepth. You wouldn't use this for a hard\nreasoning problem. You'd use it for\nbackground tasks and agent pipelines\nthat need to move fast and cheap at\nscale. Gemini 3.1 Pro. Released in\nFebruary 2026, this is the reasoning\nspecialist. Google's benchmarks claim\nroughly double the logic performance of\nthe earlier Gemini 3 Pro. It's mostly\ngated behind preview access through the\nAPI, anti-gravity, Vertex AI, and Pro or\nUltra subscriptions in the consumer app.\nIf 3.6 Flash is built for speed, 3.1 Pro\nis built for depth. The model you'd want\non a genuinely hard problem, not a quick\none. And if you're actually building on\ntop of these through the API, price is\nwhere the real-world decision gets made.\nAccording to Google's own published\nrates, 3.6 Flash runs about $1.50 per\nmillion input tokens and $7.50 per\nmillion output tokens. For context,\nthat's noticeably cheaper on the output\nside than GPT 5.6 Luna's roughly $6 per\nmillion. That gap is exactly why so many\ndevelopers default to flash tier models\nfor anything running at volume. Now, a\nhandful of specialty models worth\nknowing by name, even if we don't dwell\non each one. Nano Banana 2 is Gemini's\ncurrent image generation and editing\nmodel, replacing the older Imagen line\nentirely. And that's not a small detail\nbecause Imagen is actually shutting down\non August 17th, 2026. If you've had\nworkflows built on Imagen, that clock is\nalready running. There's also a Nano\nBanana 2 light variant built purely for\nspeed, trading a small amount of quality\nfor much faster, cheaper output at high\nvolume. VIO 3.1 is Google's video\ngeneration model, still in beta, built\nto turn a text prompt into a short clip\nwith matching audio. Gemini audio 3.5\nlive translate handles real-time\nspeech-to-speech translation across more\nthan 70 languages, already built into\nGoogle Meet and Android. And if you're\ncurious about the more niche end of the\nlineup, there's a security-focused\nvariant called 3.5 flash cyber built to\ncoordinate with vulnerability scanning\ntools. And Lyra 3.5, Google's music\nmodel, which can now generate tracks up\nto 3 minutes long from a text prompt.\nHere's the honest limitation worth\nnaming. Google ships a lot of these\nmodels fast, and the naming gets\nconfusing on purpose or not. 3.5, 3.6,\n3.1 pro, flashlight, flash cyber. If\nyou're not building on top of the API\nprofessionally, you genuinely don't need\nto memorize this list. You just need to\nknow the shape of it. Fast and cheap,\ndeep reasoning, and multimodal. That's\nreally three categories wearing a lot of\ndifferent name tags. The modes you\nactually interact with, models are the\nengine. Modes are the steering wheel.\nHere's where things get useful for\nanyone who isn't a developer. AI mode\ninside Google Search turns your search\nbar into a conversation. As of IO 2026,\nit's globally powered by Gemini 3.5\nflash and instead of 10 blue links, you\nget a written answer with follow-up\nquestions and sometimes an interactive\nwidget built on the fly. Anyone with\nsearch can use it. No subscription\nrequired. Ask something like, \"What's a\nquick dinner with what's in my fridge?\"\nand it answers in full sentences, not a\nlist of recipe blogs. Deep Think is the\nextra effort version of the Gemini app.\nIt spends more compute per answer to\nreason through harder problems\nstep-by-step. Google gates this one\nbehind Google AI Ultra and it's built\nfor genuinely difficult science or\nengineering questions, not everyday\nchat. Now, Deep Research is where this\nstops being a chatbot and starts being\nan assistant. You give it a topic and\ninstead of one reply, it plans a\nresearch strategy, opens web pages,\nreads them and if you allow it, pulls\nfrom your own Gmail and Drive, too. What\ncomes back isn't a paragraph. It's a\nfull multi-page report inside Gemini's\ncanvas. This is the part of Gemini that\nactually earns the word agent and we're\ncoming back to why that matters in a few\nminutes. Gemini Live is the voice and\ncamera mode. Say, \"Hey Google, let's\nchat\" and you're talking to it\nhands-free with the option to point your\ncamera at something and ask what it's\nlooking at live. And Canvas is the\nworkspace mode. Type, \"Create a quiz app\nabout planets\" and it writes the code,\nthe interface and the content in one\npass, right there for you to edit. One\nmore worth a mention briefly because\nit's still early. Gemini Spark, a\npersonal agent announced at IO 2026,\nmeant to run continuously in the\nbackground handling things like\nscheduling. Right now, it's limited to\nearly Ultra testers. So, treat this one\nas coming, not here. Quick gut check\nbefore we move on. If all of that sounds\nlike a lot of separate tools, that's\nfair. But, notice the pattern. Every\nsingle one of these modes is just Gemini\n3.5 or 3.6 flash wearing a different job\ntitle. You're not learning six different\nAIs, you're learning six different ways\nto ask the same brain for help.\nMultimodal in practice. Let's talk about\nwhat multimodal actually means\nday-to-day, because it's more than a\nbuzzword on a slide. Gemini reads and\nwrites text and code. Obviously, that's\nthe baseline. But drop a photo into a\nchat and ask it to caption or edit it,\nand Nano Banana handles that. Ask it to\nspeak an answer out loud, and Gemini's\naudio models generate that voice on the\nspot with actual control over tone and\npacing. Ask for a short video and VO\nbuilds one from scratch. Ask it to edit\nan existing clip, swap the sky, change\nthe style, and that's a separate tool\ncalled Gemini Omni doing frame-by-frame\nediting by voice command. Inside Google\nDocs and Sheets, the same underlying\nmodels can draft a document from your\nmeeting notes or build a spreadsheet out\nof a pile of invoices, complete with\nformulas and charts. Not just raw\nnumbers dumped into cells. In Slides,\nhand it a list of bullet points and it\ncan lay out an actual deck, not just\ntext on blank slides. And through Gemini\nLive's camera mode, you can point your\nphone at a menu in a language you don't\nspeak and get a live translation\noverlaid on what you're looking at. Or\nask it to identify an object it's\nlooking at through the lens. No typing\ninvolved. Here's the part worth\nremembering. You never pick the model.\nYou just say what you want. Make this an\ninfographic. Translate this. Write this\nin Python. And Gemini quietly roots the\nrequest to whichever model actually does\nthat job. That's the design philosophy\nin one sentence. One platform, and it\ndecides the plumbing so you don't have\nto. Where Gemini actually lives. This is\nthe part that's easy to underestimate.\nGemini isn't confined to one app. It's\nspread across nearly everything Google\nships. In Search, it's AI mode, already\ncovered. In Gmail, it's behind Smart\nCompose and auto-reply suggestions. In\nDocs, Sheets and Slides, Ultra and Pro\nPro\nPro\nget Gemini drafting text, building\nformulas, and designing slide layouts,\npulling context from your own files when\nyou let it. In Drive, it can find and\nsummarize documents for you. In Google\nMeet, it's doing live caption\ntranslation. On Android, especially\nPixel devices, it's baked straight into\nthe voice assistant. And there's a\nChrome extension that lets the browser\nsend page content straight to Gemini, so\nyou can ask questions about whatever tab\nyou're on. And for developers, all of it\nis exposed through Google AI Studio and\nthe Gemini API, plus a newer platform\ncalled antigravity for building\nmulti-agent workflows on top of it. The\nstrategic point here isn't subtle.\nGoogle isn't trying to win the best\nstandalone chatbot argument. It's trying\nto make sure you're never more than one\nproduct away from Gemini, no matter what\nyou're doing on a Google device or in a\nGoogle app. Agents, the part that\nactually matters. Now, here's the shift\nI promised earlier, the one that\nactually changes what this platform is\nfor. Everything so far has been ask a\nquestion, get an answer. Agents are\nGoogle trying to move Gemini past that\nentirely. Deep research is the clearest\nexample already live, plan, browse,\nsynthesize, write, without you\nbabysitting every step. Spark is the\nearly, still limited attempt at a\npersistent personal agent running\ncontinuously in the background. And on\nthe developer side, antigravity lets\ncompanies build coordinated teams of\nsub-agents. Google's own blog post gave\nan example of businesses running\nparallel agents to analyze data at\nscale, rather than one model doing\neverything sequentially. Picture the\ndifference in practice. The old way, you\nask Gemini, \"What should I know before a\ntrip to Japan?\" and it gives you a\nparagraph. The agent way, you say, \"Plan\nmy trip to Japan.\" and it checks\nflights, compares hotel options, and\ncomes back with an actual itinerary,\npausing to confirm with you before it\nbooks anything. That's the same\nunderlying model, just given permission\nto take more than one step before\nhanding control back to you. None of\nthis is science fiction anymore, and\nnone of it is fully finished, either.\nThat's the honest read. Deep Research\ngenuinely works today. Spark is still in\nearly testing, but the direction is\nunmistakable. Google wants Gemini to\neventually take a task, break it into\nsteps, and execute most of them without\nyou typing a follow-up for every single\none. What actually makes Gemini\ndifferent? So, how does this stack up\nagainst everyone else building the same\nkind of thing? Let's be balanced here,\nbecause Google's advantages are real,\nbut so are its weak spots. The clearest\nedge is data. Gemini can pull from live\nsearch results, Maps, Gmail, and Drive\nin ways that a closed sandbox chatbot\nsimply can't match without plugins\nbolted on. The second edge is reach.\nEvery Android phone is a potential\nGemini client, and every Workspace\nbusiness account already has it\navailable. No competitor has that kind\nof built-in distribution. And on raw\nbenchmarks, Gemini 3 Pro topped the LM\nArena leaderboard, which, regardless of\nhow much weight you put on any single\nleaderboard, says Google's\ninfrastructure and DeepMind's research\nare producing real, top-tier results,\nnot just hype. But, and this matters for\ncredibility, Google is genuinely more\nconservative about rollout than some\ncompetitors. Deep Think and Spark are\nstill gated behind ultra subscriptions\nor limited testing, while some rivals\nship new capabilities to everyone at\nonce. And the tier structure itself,\nfree, pro, ultra, API pricing, can be\ngenuinely confusing next to a simpler\nflat subscription from a competitor. If\nyou've ever opened the Gemini pricing\npage and closed it 5 minutes later still\nunsure which plan you need, that's not\njust you. Where this is actually headed,\na few things are confirmed and a few are\nstill rumor, and it's worth keeping\nthose separate. Confirmed, Gemini 3.5\nPro is currently in partner testing with\na public release expected soon, and\nGoogle has already started training on\nGemini 4, according to its own July 2026\nannouncement, though there's no public\ntimeline for that yet. Workspace AI\nrollout continues expanding, and Gemini\nLive's regional language support keeps\ngrowing. Speculative and worth labeling\nclearly as such, there's talk of a\ndedicated on-device AI chip for future\nPixel phones, and some experimental\nDeepMind research around 3D avatars and\nworld simulation that hasn't shipped as\na product. Treat both of those as\npossible, not coming. Nothing official\nhas confirmed either one. The verdict.\nSo, where does that leave things? Gemini\nin 2026 isn't a chatbot you occasionally\nopen. It's an AI layer Google has\nthreaded through search, Gmail, your\ndocuments, and increasingly your phone\nitself. The models handle the thinking,\nthe modes handle how you ask, and agents\nlike Deep Research are the clearest sign\nof where all of it is actually heading.\nIf there's one thing worth trying this\nweek, it's Deep Research on something\nyou'd normally spend an evening looking\ninto yourself, and actually watching it\nwork instead of just reading the final\nreport. Drop a comment with which piece\nof this surprised you most, the model\nlineup, the agent side, or just how much\nof this you were already using without\nrealizing it. I'll be back soon with a\ndeeper breakdown on how Deep Research\nactually performs against a real\nresearch task. Thanks for watching, and\nI'll see you in the next one.",
  "language": "en",
  "duration": 997.248,
  "segments": [
    {
      "start": 0.0,
      "end": 1.48,
      "text": "You've probably typed something into",
      "source": "source_caption"
    },
    {
      "start": 1.48,
      "end": 3.56,
      "text": "Gemini, got an answer, and closed the",
      "source": "source_caption"
    },
    {
      "start": 3.56,
      "end": 5.92,
      "text": "tab, the same way you'd use any other",
      "source": "source_caption"
    },
    {
      "start": 5.92,
      "end": 8.2,
      "text": "chatbot. Here's the thing, that's maybe",
      "source": "source_caption"
    },
    {
      "start": 8.2,
      "end": 10.84,
      "text": "10% of what it actually does. Use it",
      "source": "source_caption"
    },
    {
      "start": 10.84,
      "end": 12.88,
      "text": "right, and it can quietly take over",
      "source": "source_caption"
    },
    {
      "start": 12.88,
      "end": 14.64,
      "text": "almost half the busy work you're still",
      "source": "source_caption"
    },
    {
      "start": 14.64,
      "end": 17.04,
      "text": "doing by hand. I spent hours mapping",
      "source": "source_caption"
    },
    {
      "start": 17.04,
      "end": 19.48,
      "text": "every model, every mode, and every",
      "source": "source_caption"
    },
    {
      "start": 19.48,
      "end": 22.12,
      "text": "product Gemini is quietly wired into,",
      "source": "source_caption"
    },
    {
      "start": 22.12,
      "end": 24.16,
      "text": "and the number that stopped me was this.",
      "source": "source_caption"
    },
    {
      "start": 24.16,
      "end": 27.16,
      "text": "Gemini's app alone has over 650 million",
      "source": "source_caption"
    },
    {
      "start": 27.16,
      "end": 29.32,
      "text": "monthly users, and that's before you",
      "source": "source_caption"
    },
    {
      "start": 29.32,
      "end": 31.64,
      "text": "count everyone using it inside search.",
      "source": "source_caption"
    },
    {
      "start": 31.64,
      "end": 34.2,
      "text": "Most of them are using maybe 20% of it,",
      "source": "source_caption"
    },
    {
      "start": 34.2,
      "end": 36.68,
      "text": "with no idea the rest even exists. Look",
      "source": "source_caption"
    },
    {
      "start": 36.68,
      "end": 39.04,
      "text": "at this data from the Census Bureau.",
      "source": "source_caption"
    },
    {
      "start": 39.04,
      "end": 41.4,
      "text": "Only about one in five US businesses",
      "source": "source_caption"
    },
    {
      "start": 41.4,
      "end": 44.4,
      "text": "actually use AI in their operations.",
      "source": "source_caption"
    },
    {
      "start": 44.4,
      "end": 46.04,
      "text": "So, if you run a business, and you're",
      "source": "source_caption"
    },
    {
      "start": 46.04,
      "end": 48.08,
      "text": "even thinking about this, you're ahead",
      "source": "source_caption"
    },
    {
      "start": 48.08,
      "end": 50.4,
      "text": "of most of your competition.",
      "source": "source_caption"
    },
    {
      "start": 50.4,
      "end": 51.72,
      "text": "What you might not know is that",
      "source": "source_caption"
    },
    {
      "start": 51.72,
      "end": 54.4,
      "text": "alongside covering AI news, we work with",
      "source": "source_caption"
    },
    {
      "start": 54.4,
      "end": 55.96,
      "text": "business owners to help them implement",
      "source": "source_caption"
    },
    {
      "start": 55.96,
      "end": 57.96,
      "text": "AI in their business.",
      "source": "source_caption"
    },
    {
      "start": 57.96,
      "end": 59.52,
      "text": "Our engineering team gets to know how",
      "source": "source_caption"
    },
    {
      "start": 59.52,
      "end": 61.44,
      "text": "your business runs, then builds the",
      "source": "source_caption"
    },
    {
      "start": 61.44,
      "end": 63.08,
      "text": "automation with you.",
      "source": "source_caption"
    },
    {
      "start": 63.08,
      "end": 64.44,
      "text": "You'll find the link in the description",
      "source": "source_caption"
    },
    {
      "start": 64.44,
      "end": 65.47,
      "text": "below.",
      "source": "source_caption"
    },
    {
      "start": 65.47,
      "end": 65.48,
      "text": "below.",
      "source": "source_caption"
    },
    {
      "start": 65.48,
      "end": 67.4,
      "text": "Click it, fill out a short form about",
      "source": "source_caption"
    },
    {
      "start": 67.4,
      "end": 69.24,
      "text": "your business, and we'll get in touch to",
      "source": "source_caption"
    },
    {
      "start": 69.24,
      "end": 71.44,
      "text": "set up a call. So, in this video, I'm",
      "source": "source_caption"
    },
    {
      "start": 71.44,
      "end": 74.28,
      "text": "breaking down exactly what Gemini is as",
      "source": "source_caption"
    },
    {
      "start": 74.28,
      "end": 76.24,
      "text": "of mid-2026,",
      "source": "source_caption"
    },
    {
      "start": 76.24,
      "end": 78.96,
      "text": "every current model, every mode, and",
      "source": "source_caption"
    },
    {
      "start": 78.96,
      "end": 81.0,
      "text": "everywhere Google has quietly built it",
      "source": "source_caption"
    },
    {
      "start": 81.0,
      "end": 83.28,
      "text": "in. By the end, you'll know exactly",
      "source": "source_caption"
    },
    {
      "start": 83.28,
      "end": 85.56,
      "text": "which Gemini tool to reach for depending",
      "source": "source_caption"
    },
    {
      "start": 85.56,
      "end": 87.48,
      "text": "on what you're actually trying to do,",
      "source": "source_caption"
    },
    {
      "start": 87.48,
      "end": 89.2,
      "text": "instead of just typing into whichever",
      "source": "source_caption"
    },
    {
      "start": 89.2,
      "end": 91.32,
      "text": "box is in front of you. First, let's",
      "source": "source_caption"
    },
    {
      "start": 91.32,
      "end": 93.24,
      "text": "clear up the biggest misconception.",
      "source": "source_caption"
    },
    {
      "start": 93.24,
      "end": 95.76,
      "text": "Gemini isn't one product at all. What",
      "source": "source_caption"
    },
    {
      "start": 95.76,
      "end": 98.24,
      "text": "Gemini actually is, here's the mental",
      "source": "source_caption"
    },
    {
      "start": 98.24,
      "end": 100.04,
      "text": "model you need before any of this makes",
      "source": "source_caption"
    },
    {
      "start": 100.04,
      "end": 102.88,
      "text": "sense. Gemini isn't a single AI, it's",
      "source": "source_caption"
    },
    {
      "start": 102.88,
      "end": 104.48,
      "text": "Google's umbrella name for a whole",
      "source": "source_caption"
    },
    {
      "start": 104.48,
      "end": 107.2,
      "text": "platform, a family of models underneath,",
      "source": "source_caption"
    },
    {
      "start": 107.2,
      "end": 108.88,
      "text": "and a set of products on top that let",
      "source": "source_caption"
    },
    {
      "start": 108.88,
      "end": 110.64,
      "text": "you actually talk to them. Think of it",
      "source": "source_caption"
    },
    {
      "start": 110.64,
      "end": 112.72,
      "text": "in two layers. The bottom layer is the",
      "source": "source_caption"
    },
    {
      "start": 112.72,
      "end": 114.84,
      "text": "models themselves, things like Gemini",
      "source": "source_caption"
    },
    {
      "start": 114.84,
      "end": 118.52,
      "text": "3.6 Flash or Gemini 3.1 Pro. These are",
      "source": "source_caption"
    },
    {
      "start": 118.52,
      "end": 120.92,
      "text": "the engines tuned for different jobs.",
      "source": "source_caption"
    },
    {
      "start": 120.92,
      "end": 123.0,
      "text": "Some built for speed, some for heavy",
      "source": "source_caption"
    },
    {
      "start": 123.0,
      "end": 125.76,
      "text": "reasoning, some for images or audio. You",
      "source": "source_caption"
    },
    {
      "start": 125.76,
      "end": 127.12,
      "text": "never see these names unless you go",
      "source": "source_caption"
    },
    {
      "start": 127.12,
      "end": 129.28,
      "text": "looking. The top layer is everything you",
      "source": "source_caption"
    },
    {
      "start": 129.28,
      "end": 131.959,
      "text": "actually click on. The Gemini app, AI",
      "source": "source_caption"
    },
    {
      "start": 131.959,
      "end": 134.64,
      "text": "mode inside Google search, Gemini inside",
      "source": "source_caption"
    },
    {
      "start": 134.64,
      "end": 137.04,
      "text": "Gmail and Docs, the voice assistant on",
      "source": "source_caption"
    },
    {
      "start": 137.04,
      "end": 138.68,
      "text": "your phone. All of those are just",
      "source": "source_caption"
    },
    {
      "start": 138.68,
      "end": 140.68,
      "text": "different doors into the same underlying",
      "source": "source_caption"
    },
    {
      "start": 140.68,
      "end": 142.4,
      "text": "models. That's the whole point of this",
      "source": "source_caption"
    },
    {
      "start": 142.4,
      "end": 144.68,
      "text": "video. Google isn't trying to build one",
      "source": "source_caption"
    },
    {
      "start": 144.68,
      "end": 146.56,
      "text": "great chatbot. It's trying to put the",
      "source": "source_caption"
    },
    {
      "start": 146.56,
      "end": 148.8,
      "text": "same AI brain behind every product you",
      "source": "source_caption"
    },
    {
      "start": 148.8,
      "end": 150.68,
      "text": "already use. So, let's start with the",
      "source": "source_caption"
    },
    {
      "start": 150.68,
      "end": 153.68,
      "text": "brains, the actual models, because once",
      "source": "source_caption"
    },
    {
      "start": 153.68,
      "end": 155.76,
      "text": "you know what each one is built for,",
      "source": "source_caption"
    },
    {
      "start": 155.76,
      "end": 157.8,
      "text": "everything else clicks into place. The",
      "source": "source_caption"
    },
    {
      "start": 157.8,
      "end": 159.959,
      "text": "current model lineup. This is a demo",
      "source": "source_caption"
    },
    {
      "start": 159.959,
      "end": 162.0,
      "text": "checklist, so we're going model by",
      "source": "source_caption"
    },
    {
      "start": 162.0,
      "end": 164.32,
      "text": "model. What it is, what it's actually",
      "source": "source_caption"
    },
    {
      "start": 164.32,
      "end": 166.4,
      "text": "good for, and where you can get it.",
      "source": "source_caption"
    },
    {
      "start": 166.4,
      "end": 168.84,
      "text": "Gemini 3.7 flash.",
      "source": "source_caption"
    },
    {
      "start": 168.84,
      "end": 171.6,
      "text": "Launched on August 13th, 2026, this is",
      "source": "source_caption"
    },
    {
      "start": 171.6,
      "end": 173.44,
      "text": "Google's newest flash model and its most",
      "source": "source_caption"
    },
    {
      "start": 173.44,
      "end": 175.48,
      "text": "capable workhorse yet. It's built",
      "source": "source_caption"
    },
    {
      "start": 175.48,
      "end": 178.04,
      "text": "primarily for coding and AI agents with",
      "source": "source_caption"
    },
    {
      "start": 178.04,
      "end": 179.4,
      "text": "major improvements in software",
      "source": "source_caption"
    },
    {
      "start": 179.4,
      "end": 181.519,
      "text": "engineering, web development, and",
      "source": "source_caption"
    },
    {
      "start": 181.519,
      "end": 184.32,
      "text": "complex multi-step workflows. Google has",
      "source": "source_caption"
    },
    {
      "start": 184.32,
      "end": 185.84,
      "text": "already made it generally available",
      "source": "source_caption"
    },
    {
      "start": 185.84,
      "end": 189.12,
      "text": "through the Gemini API, positioning 3.7",
      "source": "source_caption"
    },
    {
      "start": 189.12,
      "end": 191.56,
      "text": "flash as the new go-to model when you",
      "source": "source_caption"
    },
    {
      "start": 191.56,
      "end": 193.28,
      "text": "want strong intelligence without giving",
      "source": "source_caption"
    },
    {
      "start": 193.28,
      "end": 195.08,
      "text": "up the speed and efficiency the flash",
      "source": "source_caption"
    },
    {
      "start": 195.08,
      "end": 198.32,
      "text": "lineup is known for. Gemini 3.6 flash.",
      "source": "source_caption"
    },
    {
      "start": 198.32,
      "end": 200.28,
      "text": "This is Google's current flagship,",
      "source": "source_caption"
    },
    {
      "start": 200.28,
      "end": 202.32,
      "text": "announced in a company blog post on July",
      "source": "source_caption"
    },
    {
      "start": 202.32,
      "end": 205.72,
      "text": "21st, 2026. It's built as a workhorse,",
      "source": "source_caption"
    },
    {
      "start": 205.72,
      "end": 207.92,
      "text": "strong at coding, knowledge work, and",
      "source": "source_caption"
    },
    {
      "start": 207.92,
      "end": 209.92,
      "text": "multimodal tasks. And according to",
      "source": "source_caption"
    },
    {
      "start": 209.92,
      "end": 212.08,
      "text": "Google's own numbers, it does the job",
      "source": "source_caption"
    },
    {
      "start": 212.08,
      "end": 215.12,
      "text": "using about 17% fewer tokens on average",
      "source": "source_caption"
    },
    {
      "start": 215.12,
      "end": 217.32,
      "text": "than its predecessor. Fewer tokens means",
      "source": "source_caption"
    },
    {
      "start": 217.32,
      "end": 219.519,
      "text": "faster answers and a lower bill if",
      "source": "source_caption"
    },
    {
      "start": 219.519,
      "end": 221.6,
      "text": "you're paying for it through the API.",
      "source": "source_caption"
    },
    {
      "start": 221.6,
      "end": 224.08,
      "text": "You can reach it through the Gemini API,",
      "source": "source_caption"
    },
    {
      "start": 224.08,
      "end": 226.64,
      "text": "through AI Studio, or simply by using",
      "source": "source_caption"
    },
    {
      "start": 226.64,
      "end": 229.64,
      "text": "the Gemini app and searches AI mode. No",
      "source": "source_caption"
    },
    {
      "start": 229.64,
      "end": 231.519,
      "text": "extra setup required. If you only",
      "source": "source_caption"
    },
    {
      "start": 231.519,
      "end": 233.76,
      "text": "remember one model name from this video,",
      "source": "source_caption"
    },
    {
      "start": 233.76,
      "end": 235.72,
      "text": "make it this one because it's what most",
      "source": "source_caption"
    },
    {
      "start": 235.72,
      "end": 237.56,
      "text": "of Gemini is quietly running on right",
      "source": "source_caption"
    },
    {
      "start": 237.56,
      "end": 240.96,
      "text": "now. Gemini 3.5 Flash. This one launched",
      "source": "source_caption"
    },
    {
      "start": 240.96,
      "end": 243.24,
      "text": "back in May 2026",
      "source": "source_caption"
    },
    {
      "start": 243.24,
      "end": 244.8,
      "text": "and it's the model that was actually",
      "source": "source_caption"
    },
    {
      "start": 244.8,
      "end": 248.04,
      "text": "powering AI mode in search before 3.6",
      "source": "source_caption"
    },
    {
      "start": 248.04,
      "end": 250.32,
      "text": "Flash took over. Google described it as",
      "source": "source_caption"
    },
    {
      "start": 250.32,
      "end": 251.959,
      "text": "frontier-level intelligence at",
      "source": "source_caption"
    },
    {
      "start": 251.959,
      "end": 253.88,
      "text": "exceptional speed and by its own",
      "source": "source_caption"
    },
    {
      "start": 253.88,
      "end": 256.4,
      "text": "benchmarks, it pushed output throughput",
      "source": "source_caption"
    },
    {
      "start": 256.4,
      "end": 258.359,
      "text": "to roughly four times faster than other",
      "source": "source_caption"
    },
    {
      "start": 258.359,
      "end": 260.48,
      "text": "top models at the time. It's still very",
      "source": "source_caption"
    },
    {
      "start": 260.48,
      "end": 262.44,
      "text": "much in active use across the Gemini",
      "source": "source_caption"
    },
    {
      "start": 262.44,
      "end": 265.36,
      "text": "app, Google's anti-gravity platform, and",
      "source": "source_caption"
    },
    {
      "start": 265.36,
      "end": 267.56,
      "text": "enterprise tools. It's slightly behind",
      "source": "source_caption"
    },
    {
      "start": 267.56,
      "end": 270.12,
      "text": "3.6 now, but it's the model quietly",
      "source": "source_caption"
    },
    {
      "start": 270.12,
      "end": 271.919,
      "text": "sitting behind a huge share of what",
      "source": "source_caption"
    },
    {
      "start": 271.919,
      "end": 274.24,
      "text": "shipped this year. Gemini 3.5",
      "source": "source_caption"
    },
    {
      "start": 274.24,
      "end": 276.6,
      "text": "Flashlight. Same July announcement,",
      "source": "source_caption"
    },
    {
      "start": 276.6,
      "end": 278.56,
      "text": "different job entirely. This is the",
      "source": "source_caption"
    },
    {
      "start": 278.56,
      "end": 280.88,
      "text": "stripped-down, high-throughput sibling.",
      "source": "source_caption"
    },
    {
      "start": 280.88,
      "end": 283.56,
      "text": "Google sites roughly 350 tokens per",
      "source": "source_caption"
    },
    {
      "start": 283.56,
      "end": 285.919,
      "text": "second, which is built for volume, not",
      "source": "source_caption"
    },
    {
      "start": 285.919,
      "end": 287.76,
      "text": "depth. You wouldn't use this for a hard",
      "source": "source_caption"
    },
    {
      "start": 287.76,
      "end": 289.4,
      "text": "reasoning problem. You'd use it for",
      "source": "source_caption"
    },
    {
      "start": 289.4,
      "end": 291.919,
      "text": "background tasks and agent pipelines",
      "source": "source_caption"
    },
    {
      "start": 291.919,
      "end": 293.52,
      "text": "that need to move fast and cheap at",
      "source": "source_caption"
    },
    {
      "start": 293.52,
      "end": 296.52,
      "text": "scale. Gemini 3.1 Pro. Released in",
      "source": "source_caption"
    },
    {
      "start": 296.52,
      "end": 299.16,
      "text": "February 2026, this is the reasoning",
      "source": "source_caption"
    },
    {
      "start": 299.16,
      "end": 301.36,
      "text": "specialist. Google's benchmarks claim",
      "source": "source_caption"
    },
    {
      "start": 301.36,
      "end": 303.2,
      "text": "roughly double the logic performance of",
      "source": "source_caption"
    },
    {
      "start": 303.2,
      "end": 305.52,
      "text": "the earlier Gemini 3 Pro. It's mostly",
      "source": "source_caption"
    },
    {
      "start": 305.52,
      "end": 307.48,
      "text": "gated behind preview access through the",
      "source": "source_caption"
    },
    {
      "start": 307.48,
      "end": 311.36,
      "text": "API, anti-gravity, Vertex AI, and Pro or",
      "source": "source_caption"
    },
    {
      "start": 311.36,
      "end": 313.72,
      "text": "Ultra subscriptions in the consumer app.",
      "source": "source_caption"
    },
    {
      "start": 313.72,
      "end": 317.36,
      "text": "If 3.6 Flash is built for speed, 3.1 Pro",
      "source": "source_caption"
    },
    {
      "start": 317.36,
      "end": 319.52,
      "text": "is built for depth. The model you'd want",
      "source": "source_caption"
    },
    {
      "start": 319.52,
      "end": 321.76,
      "text": "on a genuinely hard problem, not a quick",
      "source": "source_caption"
    },
    {
      "start": 321.76,
      "end": 323.48,
      "text": "one. And if you're actually building on",
      "source": "source_caption"
    },
    {
      "start": 323.48,
      "end": 325.88,
      "text": "top of these through the API, price is",
      "source": "source_caption"
    },
    {
      "start": 325.88,
      "end": 327.96,
      "text": "where the real-world decision gets made.",
      "source": "source_caption"
    },
    {
      "start": 327.96,
      "end": 329.44,
      "text": "According to Google's own published",
      "source": "source_caption"
    },
    {
      "start": 329.44,
      "end": 333.0,
      "text": "rates, 3.6 Flash runs about $1.50 per",
      "source": "source_caption"
    },
    {
      "start": 333.0,
      "end": 335.6,
      "text": "million input tokens and $7.50 per",
      "source": "source_caption"
    },
    {
      "start": 335.6,
      "end": 338.24,
      "text": "million output tokens. For context,",
      "source": "source_caption"
    },
    {
      "start": 338.24,
      "end": 340.16,
      "text": "that's noticeably cheaper on the output",
      "source": "source_caption"
    },
    {
      "start": 340.16,
      "end": 344.32,
      "text": "side than GPT 5.6 Luna's roughly $6 per",
      "source": "source_caption"
    },
    {
      "start": 344.32,
      "end": 347.0,
      "text": "million. That gap is exactly why so many",
      "source": "source_caption"
    },
    {
      "start": 347.0,
      "end": 349.48,
      "text": "developers default to flash tier models",
      "source": "source_caption"
    },
    {
      "start": 349.48,
      "end": 351.96,
      "text": "for anything running at volume. Now, a",
      "source": "source_caption"
    },
    {
      "start": 351.96,
      "end": 353.68,
      "text": "handful of specialty models worth",
      "source": "source_caption"
    },
    {
      "start": 353.68,
      "end": 355.92,
      "text": "knowing by name, even if we don't dwell",
      "source": "source_caption"
    },
    {
      "start": 355.92,
      "end": 358.68,
      "text": "on each one. Nano Banana 2 is Gemini's",
      "source": "source_caption"
    },
    {
      "start": 358.68,
      "end": 360.72,
      "text": "current image generation and editing",
      "source": "source_caption"
    },
    {
      "start": 360.72,
      "end": 363.2,
      "text": "model, replacing the older Imagen line",
      "source": "source_caption"
    },
    {
      "start": 363.2,
      "end": 366.0,
      "text": "entirely. And that's not a small detail",
      "source": "source_caption"
    },
    {
      "start": 366.0,
      "end": 367.88,
      "text": "because Imagen is actually shutting down",
      "source": "source_caption"
    },
    {
      "start": 367.88,
      "end": 370.92,
      "text": "on August 17th, 2026. If you've had",
      "source": "source_caption"
    },
    {
      "start": 370.92,
      "end": 373.28,
      "text": "workflows built on Imagen, that clock is",
      "source": "source_caption"
    },
    {
      "start": 373.28,
      "end": 375.2,
      "text": "already running. There's also a Nano",
      "source": "source_caption"
    },
    {
      "start": 375.2,
      "end": 377.44,
      "text": "Banana 2 light variant built purely for",
      "source": "source_caption"
    },
    {
      "start": 377.44,
      "end": 379.919,
      "text": "speed, trading a small amount of quality",
      "source": "source_caption"
    },
    {
      "start": 379.919,
      "end": 382.0,
      "text": "for much faster, cheaper output at high",
      "source": "source_caption"
    },
    {
      "start": 382.0,
      "end": 384.64,
      "text": "volume. VIO 3.1 is Google's video",
      "source": "source_caption"
    },
    {
      "start": 384.64,
      "end": 387.12,
      "text": "generation model, still in beta, built",
      "source": "source_caption"
    },
    {
      "start": 387.12,
      "end": 389.16,
      "text": "to turn a text prompt into a short clip",
      "source": "source_caption"
    },
    {
      "start": 389.16,
      "end": 392.36,
      "text": "with matching audio. Gemini audio 3.5",
      "source": "source_caption"
    },
    {
      "start": 392.36,
      "end": 394.36,
      "text": "live translate handles real-time",
      "source": "source_caption"
    },
    {
      "start": 394.36,
      "end": 396.919,
      "text": "speech-to-speech translation across more",
      "source": "source_caption"
    },
    {
      "start": 396.919,
      "end": 399.36,
      "text": "than 70 languages, already built into",
      "source": "source_caption"
    },
    {
      "start": 399.36,
      "end": 401.16,
      "text": "Google Meet and Android. And if you're",
      "source": "source_caption"
    },
    {
      "start": 401.16,
      "end": 402.919,
      "text": "curious about the more niche end of the",
      "source": "source_caption"
    },
    {
      "start": 402.919,
      "end": 405.08,
      "text": "lineup, there's a security-focused",
      "source": "source_caption"
    },
    {
      "start": 405.08,
      "end": 408.2,
      "text": "variant called 3.5 flash cyber built to",
      "source": "source_caption"
    },
    {
      "start": 408.2,
      "end": 410.0,
      "text": "coordinate with vulnerability scanning",
      "source": "source_caption"
    },
    {
      "start": 410.0,
      "end": 413.52,
      "text": "tools. And Lyra 3.5, Google's music",
      "source": "source_caption"
    },
    {
      "start": 413.52,
      "end": 415.72,
      "text": "model, which can now generate tracks up",
      "source": "source_caption"
    },
    {
      "start": 415.72,
      "end": 418.0,
      "text": "to 3 minutes long from a text prompt.",
      "source": "source_caption"
    },
    {
      "start": 418.0,
      "end": 419.6,
      "text": "Here's the honest limitation worth",
      "source": "source_caption"
    },
    {
      "start": 419.6,
      "end": 421.56,
      "text": "naming. Google ships a lot of these",
      "source": "source_caption"
    },
    {
      "start": 421.56,
      "end": 423.52,
      "text": "models fast, and the naming gets",
      "source": "source_caption"
    },
    {
      "start": 423.52,
      "end": 427.56,
      "text": "confusing on purpose or not. 3.5, 3.6,",
      "source": "source_caption"
    },
    {
      "start": 427.56,
      "end": 431.16,
      "text": "3.1 pro, flashlight, flash cyber. If",
      "source": "source_caption"
    },
    {
      "start": 431.16,
      "end": 432.96,
      "text": "you're not building on top of the API",
      "source": "source_caption"
    },
    {
      "start": 432.96,
      "end": 435.32,
      "text": "professionally, you genuinely don't need",
      "source": "source_caption"
    },
    {
      "start": 435.32,
      "end": 437.48,
      "text": "to memorize this list. You just need to",
      "source": "source_caption"
    },
    {
      "start": 437.48,
      "end": 439.72,
      "text": "know the shape of it. Fast and cheap,",
      "source": "source_caption"
    },
    {
      "start": 439.72,
      "end": 442.0,
      "text": "deep reasoning, and multimodal. That's",
      "source": "source_caption"
    },
    {
      "start": 442.0,
      "end": 443.88,
      "text": "really three categories wearing a lot of",
      "source": "source_caption"
    },
    {
      "start": 443.88,
      "end": 446.04,
      "text": "different name tags. The modes you",
      "source": "source_caption"
    },
    {
      "start": 446.04,
      "end": 448.4,
      "text": "actually interact with, models are the",
      "source": "source_caption"
    },
    {
      "start": 448.4,
      "end": 450.6,
      "text": "engine. Modes are the steering wheel.",
      "source": "source_caption"
    },
    {
      "start": 450.6,
      "end": 452.08,
      "text": "Here's where things get useful for",
      "source": "source_caption"
    },
    {
      "start": 452.08,
      "end": 454.76,
      "text": "anyone who isn't a developer. AI mode",
      "source": "source_caption"
    },
    {
      "start": 454.76,
      "end": 456.6,
      "text": "inside Google Search turns your search",
      "source": "source_caption"
    },
    {
      "start": 456.6,
      "end": 460.36,
      "text": "bar into a conversation. As of IO 2026,",
      "source": "source_caption"
    },
    {
      "start": 460.36,
      "end": 462.96,
      "text": "it's globally powered by Gemini 3.5",
      "source": "source_caption"
    },
    {
      "start": 462.96,
      "end": 465.56,
      "text": "flash and instead of 10 blue links, you",
      "source": "source_caption"
    },
    {
      "start": 465.56,
      "end": 467.04,
      "text": "get a written answer with follow-up",
      "source": "source_caption"
    },
    {
      "start": 467.04,
      "end": 469.24,
      "text": "questions and sometimes an interactive",
      "source": "source_caption"
    },
    {
      "start": 469.24,
      "end": 471.16,
      "text": "widget built on the fly. Anyone with",
      "source": "source_caption"
    },
    {
      "start": 471.16,
      "end": 473.0,
      "text": "search can use it. No subscription",
      "source": "source_caption"
    },
    {
      "start": 473.0,
      "end": 475.24,
      "text": "required. Ask something like, \"What's a",
      "source": "source_caption"
    },
    {
      "start": 475.24,
      "end": 477.12,
      "text": "quick dinner with what's in my fridge?\"",
      "source": "source_caption"
    },
    {
      "start": 477.12,
      "end": 479.32,
      "text": "and it answers in full sentences, not a",
      "source": "source_caption"
    },
    {
      "start": 479.32,
      "end": 481.72,
      "text": "list of recipe blogs. Deep Think is the",
      "source": "source_caption"
    },
    {
      "start": 481.72,
      "end": 484.08,
      "text": "extra effort version of the Gemini app.",
      "source": "source_caption"
    },
    {
      "start": 484.08,
      "end": 486.12,
      "text": "It spends more compute per answer to",
      "source": "source_caption"
    },
    {
      "start": 486.12,
      "end": 487.84,
      "text": "reason through harder problems",
      "source": "source_caption"
    },
    {
      "start": 487.84,
      "end": 490.08,
      "text": "step-by-step. Google gates this one",
      "source": "source_caption"
    },
    {
      "start": 490.08,
      "end": 492.28,
      "text": "behind Google AI Ultra and it's built",
      "source": "source_caption"
    },
    {
      "start": 492.28,
      "end": 494.16,
      "text": "for genuinely difficult science or",
      "source": "source_caption"
    },
    {
      "start": 494.16,
      "end": 496.2,
      "text": "engineering questions, not everyday",
      "source": "source_caption"
    },
    {
      "start": 496.2,
      "end": 498.76,
      "text": "chat. Now, Deep Research is where this",
      "source": "source_caption"
    },
    {
      "start": 498.76,
      "end": 500.84,
      "text": "stops being a chatbot and starts being",
      "source": "source_caption"
    },
    {
      "start": 500.84,
      "end": 502.96,
      "text": "an assistant. You give it a topic and",
      "source": "source_caption"
    },
    {
      "start": 502.96,
      "end": 504.96,
      "text": "instead of one reply, it plans a",
      "source": "source_caption"
    },
    {
      "start": 504.96,
      "end": 507.68,
      "text": "research strategy, opens web pages,",
      "source": "source_caption"
    },
    {
      "start": 507.68,
      "end": 510.6,
      "text": "reads them and if you allow it, pulls",
      "source": "source_caption"
    },
    {
      "start": 510.6,
      "end": 512.8,
      "text": "from your own Gmail and Drive, too. What",
      "source": "source_caption"
    },
    {
      "start": 512.8,
      "end": 514.84,
      "text": "comes back isn't a paragraph. It's a",
      "source": "source_caption"
    },
    {
      "start": 514.84,
      "end": 517.32,
      "text": "full multi-page report inside Gemini's",
      "source": "source_caption"
    },
    {
      "start": 517.32,
      "end": 519.76,
      "text": "canvas. This is the part of Gemini that",
      "source": "source_caption"
    },
    {
      "start": 519.76,
      "end": 522.479,
      "text": "actually earns the word agent and we're",
      "source": "source_caption"
    },
    {
      "start": 522.479,
      "end": 524.32,
      "text": "coming back to why that matters in a few",
      "source": "source_caption"
    },
    {
      "start": 524.32,
      "end": 526.64,
      "text": "minutes. Gemini Live is the voice and",
      "source": "source_caption"
    },
    {
      "start": 526.64,
      "end": 529.12,
      "text": "camera mode. Say, \"Hey Google, let's",
      "source": "source_caption"
    },
    {
      "start": 529.12,
      "end": 530.56,
      "text": "chat\" and you're talking to it",
      "source": "source_caption"
    },
    {
      "start": 530.56,
      "end": 532.64,
      "text": "hands-free with the option to point your",
      "source": "source_caption"
    },
    {
      "start": 532.64,
      "end": 534.36,
      "text": "camera at something and ask what it's",
      "source": "source_caption"
    },
    {
      "start": 534.36,
      "end": 536.72,
      "text": "looking at live. And Canvas is the",
      "source": "source_caption"
    },
    {
      "start": 536.72,
      "end": 539.32,
      "text": "workspace mode. Type, \"Create a quiz app",
      "source": "source_caption"
    },
    {
      "start": 539.32,
      "end": 541.72,
      "text": "about planets\" and it writes the code,",
      "source": "source_caption"
    },
    {
      "start": 541.72,
      "end": 543.72,
      "text": "the interface and the content in one",
      "source": "source_caption"
    },
    {
      "start": 543.72,
      "end": 546.24,
      "text": "pass, right there for you to edit. One",
      "source": "source_caption"
    },
    {
      "start": 546.24,
      "end": 548.32,
      "text": "more worth a mention briefly because",
      "source": "source_caption"
    },
    {
      "start": 548.32,
      "end": 550.68,
      "text": "it's still early. Gemini Spark, a",
      "source": "source_caption"
    },
    {
      "start": 550.68,
      "end": 554.08,
      "text": "personal agent announced at IO 2026,",
      "source": "source_caption"
    },
    {
      "start": 554.08,
      "end": 555.52,
      "text": "meant to run continuously in the",
      "source": "source_caption"
    },
    {
      "start": 555.52,
      "end": 556.96,
      "text": "background handling things like",
      "source": "source_caption"
    },
    {
      "start": 556.96,
      "end": 559.0,
      "text": "scheduling. Right now, it's limited to",
      "source": "source_caption"
    },
    {
      "start": 559.0,
      "end": 561.28,
      "text": "early Ultra testers. So, treat this one",
      "source": "source_caption"
    },
    {
      "start": 561.28,
      "end": 563.68,
      "text": "as coming, not here. Quick gut check",
      "source": "source_caption"
    },
    {
      "start": 563.68,
      "end": 565.88,
      "text": "before we move on. If all of that sounds",
      "source": "source_caption"
    },
    {
      "start": 565.88,
      "end": 567.72,
      "text": "like a lot of separate tools, that's",
      "source": "source_caption"
    },
    {
      "start": 567.72,
      "end": 569.88,
      "text": "fair. But, notice the pattern. Every",
      "source": "source_caption"
    },
    {
      "start": 569.88,
      "end": 571.96,
      "text": "single one of these modes is just Gemini",
      "source": "source_caption"
    },
    {
      "start": 571.96,
      "end": 575.16,
      "text": "3.5 or 3.6 flash wearing a different job",
      "source": "source_caption"
    },
    {
      "start": 575.16,
      "end": 577.24,
      "text": "title. You're not learning six different",
      "source": "source_caption"
    },
    {
      "start": 577.24,
      "end": 579.56,
      "text": "AIs, you're learning six different ways",
      "source": "source_caption"
    },
    {
      "start": 579.56,
      "end": 581.4,
      "text": "to ask the same brain for help.",
      "source": "source_caption"
    },
    {
      "start": 581.4,
      "end": 583.72,
      "text": "Multimodal in practice. Let's talk about",
      "source": "source_caption"
    },
    {
      "start": 583.72,
      "end": 585.4,
      "text": "what multimodal actually means",
      "source": "source_caption"
    },
    {
      "start": 585.4,
      "end": 587.0,
      "text": "day-to-day, because it's more than a",
      "source": "source_caption"
    },
    {
      "start": 587.0,
      "end": 589.36,
      "text": "buzzword on a slide. Gemini reads and",
      "source": "source_caption"
    },
    {
      "start": 589.36,
      "end": 591.72,
      "text": "writes text and code. Obviously, that's",
      "source": "source_caption"
    },
    {
      "start": 591.72,
      "end": 593.68,
      "text": "the baseline. But drop a photo into a",
      "source": "source_caption"
    },
    {
      "start": 593.68,
      "end": 596.12,
      "text": "chat and ask it to caption or edit it,",
      "source": "source_caption"
    },
    {
      "start": 596.12,
      "end": 598.24,
      "text": "and Nano Banana handles that. Ask it to",
      "source": "source_caption"
    },
    {
      "start": 598.24,
      "end": 600.48,
      "text": "speak an answer out loud, and Gemini's",
      "source": "source_caption"
    },
    {
      "start": 600.48,
      "end": 602.4,
      "text": "audio models generate that voice on the",
      "source": "source_caption"
    },
    {
      "start": 602.4,
      "end": 605.04,
      "text": "spot with actual control over tone and",
      "source": "source_caption"
    },
    {
      "start": 605.04,
      "end": 607.68,
      "text": "pacing. Ask for a short video and VO",
      "source": "source_caption"
    },
    {
      "start": 607.68,
      "end": 609.88,
      "text": "builds one from scratch. Ask it to edit",
      "source": "source_caption"
    },
    {
      "start": 609.88,
      "end": 612.52,
      "text": "an existing clip, swap the sky, change",
      "source": "source_caption"
    },
    {
      "start": 612.52,
      "end": 614.48,
      "text": "the style, and that's a separate tool",
      "source": "source_caption"
    },
    {
      "start": 614.48,
      "end": 617.08,
      "text": "called Gemini Omni doing frame-by-frame",
      "source": "source_caption"
    },
    {
      "start": 617.08,
      "end": 619.36,
      "text": "editing by voice command. Inside Google",
      "source": "source_caption"
    },
    {
      "start": 619.36,
      "end": 621.76,
      "text": "Docs and Sheets, the same underlying",
      "source": "source_caption"
    },
    {
      "start": 621.76,
      "end": 623.56,
      "text": "models can draft a document from your",
      "source": "source_caption"
    },
    {
      "start": 623.56,
      "end": 625.88,
      "text": "meeting notes or build a spreadsheet out",
      "source": "source_caption"
    },
    {
      "start": 625.88,
      "end": 628.16,
      "text": "of a pile of invoices, complete with",
      "source": "source_caption"
    },
    {
      "start": 628.16,
      "end": 630.4,
      "text": "formulas and charts. Not just raw",
      "source": "source_caption"
    },
    {
      "start": 630.4,
      "end": 633.16,
      "text": "numbers dumped into cells. In Slides,",
      "source": "source_caption"
    },
    {
      "start": 633.16,
      "end": 635.04,
      "text": "hand it a list of bullet points and it",
      "source": "source_caption"
    },
    {
      "start": 635.04,
      "end": 637.28,
      "text": "can lay out an actual deck, not just",
      "source": "source_caption"
    },
    {
      "start": 637.28,
      "end": 639.56,
      "text": "text on blank slides. And through Gemini",
      "source": "source_caption"
    },
    {
      "start": 639.56,
      "end": 641.56,
      "text": "Live's camera mode, you can point your",
      "source": "source_caption"
    },
    {
      "start": 641.56,
      "end": 643.52,
      "text": "phone at a menu in a language you don't",
      "source": "source_caption"
    },
    {
      "start": 643.52,
      "end": 645.4,
      "text": "speak and get a live translation",
      "source": "source_caption"
    },
    {
      "start": 645.4,
      "end": 647.28,
      "text": "overlaid on what you're looking at. Or",
      "source": "source_caption"
    },
    {
      "start": 647.28,
      "end": 649.08,
      "text": "ask it to identify an object it's",
      "source": "source_caption"
    },
    {
      "start": 649.08,
      "end": 651.12,
      "text": "looking at through the lens. No typing",
      "source": "source_caption"
    },
    {
      "start": 651.12,
      "end": 652.72,
      "text": "involved. Here's the part worth",
      "source": "source_caption"
    },
    {
      "start": 652.72,
      "end": 655.04,
      "text": "remembering. You never pick the model.",
      "source": "source_caption"
    },
    {
      "start": 655.04,
      "end": 656.92,
      "text": "You just say what you want. Make this an",
      "source": "source_caption"
    },
    {
      "start": 656.92,
      "end": 659.72,
      "text": "infographic. Translate this. Write this",
      "source": "source_caption"
    },
    {
      "start": 659.72,
      "end": 662.2,
      "text": "in Python. And Gemini quietly roots the",
      "source": "source_caption"
    },
    {
      "start": 662.2,
      "end": 664.28,
      "text": "request to whichever model actually does",
      "source": "source_caption"
    },
    {
      "start": 664.28,
      "end": 666.48,
      "text": "that job. That's the design philosophy",
      "source": "source_caption"
    },
    {
      "start": 666.48,
      "end": 669.08,
      "text": "in one sentence. One platform, and it",
      "source": "source_caption"
    },
    {
      "start": 669.08,
      "end": 670.88,
      "text": "decides the plumbing so you don't have",
      "source": "source_caption"
    },
    {
      "start": 670.88,
      "end": 673.48,
      "text": "to. Where Gemini actually lives. This is",
      "source": "source_caption"
    },
    {
      "start": 673.48,
      "end": 675.56,
      "text": "the part that's easy to underestimate.",
      "source": "source_caption"
    },
    {
      "start": 675.56,
      "end": 678.12,
      "text": "Gemini isn't confined to one app. It's",
      "source": "source_caption"
    },
    {
      "start": 678.12,
      "end": 679.839,
      "text": "spread across nearly everything Google",
      "source": "source_caption"
    },
    {
      "start": 679.839,
      "end": 682.56,
      "text": "ships. In Search, it's AI mode, already",
      "source": "source_caption"
    },
    {
      "start": 682.56,
      "end": 684.96,
      "text": "covered. In Gmail, it's behind Smart",
      "source": "source_caption"
    },
    {
      "start": 684.96,
      "end": 687.44,
      "text": "Compose and auto-reply suggestions. In",
      "source": "source_caption"
    },
    {
      "start": 687.44,
      "end": 690.0,
      "text": "Docs, Sheets and Slides, Ultra and Pro",
      "source": "source_caption"
    },
    {
      "start": 690.0,
      "end": 690.71,
      "text": "Pro",
      "source": "source_caption"
    },
    {
      "start": 690.71,
      "end": 690.72,
      "text": "Pro",
      "source": "source_caption"
    },
    {
      "start": 690.72,
      "end": 693.0,
      "text": "get Gemini drafting text, building",
      "source": "source_caption"
    },
    {
      "start": 693.0,
      "end": 695.84,
      "text": "formulas, and designing slide layouts,",
      "source": "source_caption"
    },
    {
      "start": 695.84,
      "end": 697.8,
      "text": "pulling context from your own files when",
      "source": "source_caption"
    },
    {
      "start": 697.8,
      "end": 700.12,
      "text": "you let it. In Drive, it can find and",
      "source": "source_caption"
    },
    {
      "start": 700.12,
      "end": 702.28,
      "text": "summarize documents for you. In Google",
      "source": "source_caption"
    },
    {
      "start": 702.28,
      "end": 703.76,
      "text": "Meet, it's doing live caption",
      "source": "source_caption"
    },
    {
      "start": 703.76,
      "end": 706.12,
      "text": "translation. On Android, especially",
      "source": "source_caption"
    },
    {
      "start": 706.12,
      "end": 708.36,
      "text": "Pixel devices, it's baked straight into",
      "source": "source_caption"
    },
    {
      "start": 708.36,
      "end": 709.92,
      "text": "the voice assistant. And there's a",
      "source": "source_caption"
    },
    {
      "start": 709.92,
      "end": 711.68,
      "text": "Chrome extension that lets the browser",
      "source": "source_caption"
    },
    {
      "start": 711.68,
      "end": 714.08,
      "text": "send page content straight to Gemini, so",
      "source": "source_caption"
    },
    {
      "start": 714.08,
      "end": 716.0,
      "text": "you can ask questions about whatever tab",
      "source": "source_caption"
    },
    {
      "start": 716.0,
      "end": 718.28,
      "text": "you're on. And for developers, all of it",
      "source": "source_caption"
    },
    {
      "start": 718.28,
      "end": 720.68,
      "text": "is exposed through Google AI Studio and",
      "source": "source_caption"
    },
    {
      "start": 720.68,
      "end": 723.28,
      "text": "the Gemini API, plus a newer platform",
      "source": "source_caption"
    },
    {
      "start": 723.28,
      "end": 725.12,
      "text": "called antigravity for building",
      "source": "source_caption"
    },
    {
      "start": 725.12,
      "end": 727.32,
      "text": "multi-agent workflows on top of it. The",
      "source": "source_caption"
    },
    {
      "start": 727.32,
      "end": 729.36,
      "text": "strategic point here isn't subtle.",
      "source": "source_caption"
    },
    {
      "start": 729.36,
      "end": 730.92,
      "text": "Google isn't trying to win the best",
      "source": "source_caption"
    },
    {
      "start": 730.92,
      "end": 733.08,
      "text": "standalone chatbot argument. It's trying",
      "source": "source_caption"
    },
    {
      "start": 733.08,
      "end": 734.6,
      "text": "to make sure you're never more than one",
      "source": "source_caption"
    },
    {
      "start": 734.6,
      "end": 736.88,
      "text": "product away from Gemini, no matter what",
      "source": "source_caption"
    },
    {
      "start": 736.88,
      "end": 738.72,
      "text": "you're doing on a Google device or in a",
      "source": "source_caption"
    },
    {
      "start": 738.72,
      "end": 741.16,
      "text": "Google app. Agents, the part that",
      "source": "source_caption"
    },
    {
      "start": 741.16,
      "end": 743.28,
      "text": "actually matters. Now, here's the shift",
      "source": "source_caption"
    },
    {
      "start": 743.28,
      "end": 744.96,
      "text": "I promised earlier, the one that",
      "source": "source_caption"
    },
    {
      "start": 744.96,
      "end": 746.88,
      "text": "actually changes what this platform is",
      "source": "source_caption"
    },
    {
      "start": 746.88,
      "end": 749.6,
      "text": "for. Everything so far has been ask a",
      "source": "source_caption"
    },
    {
      "start": 749.6,
      "end": 751.96,
      "text": "question, get an answer. Agents are",
      "source": "source_caption"
    },
    {
      "start": 751.96,
      "end": 753.88,
      "text": "Google trying to move Gemini past that",
      "source": "source_caption"
    },
    {
      "start": 753.88,
      "end": 756.28,
      "text": "entirely. Deep research is the clearest",
      "source": "source_caption"
    },
    {
      "start": 756.28,
      "end": 759.28,
      "text": "example already live, plan, browse,",
      "source": "source_caption"
    },
    {
      "start": 759.28,
      "end": 761.36,
      "text": "synthesize, write, without you",
      "source": "source_caption"
    },
    {
      "start": 761.36,
      "end": 763.72,
      "text": "babysitting every step. Spark is the",
      "source": "source_caption"
    },
    {
      "start": 763.72,
      "end": 765.48,
      "text": "early, still limited attempt at a",
      "source": "source_caption"
    },
    {
      "start": 765.48,
      "end": 767.36,
      "text": "persistent personal agent running",
      "source": "source_caption"
    },
    {
      "start": 767.36,
      "end": 769.48,
      "text": "continuously in the background. And on",
      "source": "source_caption"
    },
    {
      "start": 769.48,
      "end": 771.56,
      "text": "the developer side, antigravity lets",
      "source": "source_caption"
    },
    {
      "start": 771.56,
      "end": 773.44,
      "text": "companies build coordinated teams of",
      "source": "source_caption"
    },
    {
      "start": 773.44,
      "end": 775.839,
      "text": "sub-agents. Google's own blog post gave",
      "source": "source_caption"
    },
    {
      "start": 775.839,
      "end": 777.36,
      "text": "an example of businesses running",
      "source": "source_caption"
    },
    {
      "start": 777.36,
      "end": 779.36,
      "text": "parallel agents to analyze data at",
      "source": "source_caption"
    },
    {
      "start": 779.36,
      "end": 781.6,
      "text": "scale, rather than one model doing",
      "source": "source_caption"
    },
    {
      "start": 781.6,
      "end": 783.44,
      "text": "everything sequentially. Picture the",
      "source": "source_caption"
    },
    {
      "start": 783.44,
      "end": 785.96,
      "text": "difference in practice. The old way, you",
      "source": "source_caption"
    },
    {
      "start": 785.96,
      "end": 788.12,
      "text": "ask Gemini, \"What should I know before a",
      "source": "source_caption"
    },
    {
      "start": 788.12,
      "end": 789.8,
      "text": "trip to Japan?\" and it gives you a",
      "source": "source_caption"
    },
    {
      "start": 789.8,
      "end": 792.6,
      "text": "paragraph. The agent way, you say, \"Plan",
      "source": "source_caption"
    },
    {
      "start": 792.6,
      "end": 794.56,
      "text": "my trip to Japan.\" and it checks",
      "source": "source_caption"
    },
    {
      "start": 794.56,
      "end": 797.08,
      "text": "flights, compares hotel options, and",
      "source": "source_caption"
    },
    {
      "start": 797.08,
      "end": 799.24,
      "text": "comes back with an actual itinerary,",
      "source": "source_caption"
    },
    {
      "start": 799.24,
      "end": 800.88,
      "text": "pausing to confirm with you before it",
      "source": "source_caption"
    },
    {
      "start": 800.88,
      "end": 802.56,
      "text": "books anything. That's the same",
      "source": "source_caption"
    },
    {
      "start": 802.56,
      "end": 804.64,
      "text": "underlying model, just given permission",
      "source": "source_caption"
    },
    {
      "start": 804.64,
      "end": 806.4,
      "text": "to take more than one step before",
      "source": "source_caption"
    },
    {
      "start": 806.4,
      "end": 808.24,
      "text": "handing control back to you. None of",
      "source": "source_caption"
    },
    {
      "start": 808.24,
      "end": 810.12,
      "text": "this is science fiction anymore, and",
      "source": "source_caption"
    },
    {
      "start": 810.12,
      "end": 811.96,
      "text": "none of it is fully finished, either.",
      "source": "source_caption"
    },
    {
      "start": 811.96,
      "end": 814.0,
      "text": "That's the honest read. Deep Research",
      "source": "source_caption"
    },
    {
      "start": 814.0,
      "end": 816.44,
      "text": "genuinely works today. Spark is still in",
      "source": "source_caption"
    },
    {
      "start": 816.44,
      "end": 818.52,
      "text": "early testing, but the direction is",
      "source": "source_caption"
    },
    {
      "start": 818.52,
      "end": 820.8,
      "text": "unmistakable. Google wants Gemini to",
      "source": "source_caption"
    },
    {
      "start": 820.8,
      "end": 822.96,
      "text": "eventually take a task, break it into",
      "source": "source_caption"
    },
    {
      "start": 822.96,
      "end": 825.52,
      "text": "steps, and execute most of them without",
      "source": "source_caption"
    },
    {
      "start": 825.52,
      "end": 827.4,
      "text": "you typing a follow-up for every single",
      "source": "source_caption"
    },
    {
      "start": 827.4,
      "end": 829.44,
      "text": "one. What actually makes Gemini",
      "source": "source_caption"
    },
    {
      "start": 829.44,
      "end": 831.28,
      "text": "different? So, how does this stack up",
      "source": "source_caption"
    },
    {
      "start": 831.28,
      "end": 833.0,
      "text": "against everyone else building the same",
      "source": "source_caption"
    },
    {
      "start": 833.0,
      "end": 835.56,
      "text": "kind of thing? Let's be balanced here,",
      "source": "source_caption"
    },
    {
      "start": 835.56,
      "end": 837.52,
      "text": "because Google's advantages are real,",
      "source": "source_caption"
    },
    {
      "start": 837.52,
      "end": 839.72,
      "text": "but so are its weak spots. The clearest",
      "source": "source_caption"
    },
    {
      "start": 839.72,
      "end": 842.12,
      "text": "edge is data. Gemini can pull from live",
      "source": "source_caption"
    },
    {
      "start": 842.12,
      "end": 845.44,
      "text": "search results, Maps, Gmail, and Drive",
      "source": "source_caption"
    },
    {
      "start": 845.44,
      "end": 848.2,
      "text": "in ways that a closed sandbox chatbot",
      "source": "source_caption"
    },
    {
      "start": 848.2,
      "end": 849.96,
      "text": "simply can't match without plugins",
      "source": "source_caption"
    },
    {
      "start": 849.96,
      "end": 852.44,
      "text": "bolted on. The second edge is reach.",
      "source": "source_caption"
    },
    {
      "start": 852.44,
      "end": 854.28,
      "text": "Every Android phone is a potential",
      "source": "source_caption"
    },
    {
      "start": 854.28,
      "end": 856.48,
      "text": "Gemini client, and every Workspace",
      "source": "source_caption"
    },
    {
      "start": 856.48,
      "end": 858.08,
      "text": "business account already has it",
      "source": "source_caption"
    },
    {
      "start": 858.08,
      "end": 860.52,
      "text": "available. No competitor has that kind",
      "source": "source_caption"
    },
    {
      "start": 860.52,
      "end": 862.48,
      "text": "of built-in distribution. And on raw",
      "source": "source_caption"
    },
    {
      "start": 862.48,
      "end": 865.24,
      "text": "benchmarks, Gemini 3 Pro topped the LM",
      "source": "source_caption"
    },
    {
      "start": 865.24,
      "end": 867.68,
      "text": "Arena leaderboard, which, regardless of",
      "source": "source_caption"
    },
    {
      "start": 867.68,
      "end": 869.36,
      "text": "how much weight you put on any single",
      "source": "source_caption"
    },
    {
      "start": 869.36,
      "end": 870.839,
      "text": "leaderboard, says Google's",
      "source": "source_caption"
    },
    {
      "start": 870.839,
      "end": 873.0,
      "text": "infrastructure and DeepMind's research",
      "source": "source_caption"
    },
    {
      "start": 873.0,
      "end": 875.48,
      "text": "are producing real, top-tier results,",
      "source": "source_caption"
    },
    {
      "start": 875.48,
      "end": 878.0,
      "text": "not just hype. But, and this matters for",
      "source": "source_caption"
    },
    {
      "start": 878.0,
      "end": 880.32,
      "text": "credibility, Google is genuinely more",
      "source": "source_caption"
    },
    {
      "start": 880.32,
      "end": 881.92,
      "text": "conservative about rollout than some",
      "source": "source_caption"
    },
    {
      "start": 881.92,
      "end": 884.04,
      "text": "competitors. Deep Think and Spark are",
      "source": "source_caption"
    },
    {
      "start": 884.04,
      "end": 886.0,
      "text": "still gated behind ultra subscriptions",
      "source": "source_caption"
    },
    {
      "start": 886.0,
      "end": 888.04,
      "text": "or limited testing, while some rivals",
      "source": "source_caption"
    },
    {
      "start": 888.04,
      "end": 889.68,
      "text": "ship new capabilities to everyone at",
      "source": "source_caption"
    },
    {
      "start": 889.68,
      "end": 892.2,
      "text": "once. And the tier structure itself,",
      "source": "source_caption"
    },
    {
      "start": 892.2,
      "end": 896.56,
      "text": "free, pro, ultra, API pricing, can be",
      "source": "source_caption"
    },
    {
      "start": 896.56,
      "end": 898.48,
      "text": "genuinely confusing next to a simpler",
      "source": "source_caption"
    },
    {
      "start": 898.48,
      "end": 900.6,
      "text": "flat subscription from a competitor. If",
      "source": "source_caption"
    },
    {
      "start": 900.6,
      "end": 902.12,
      "text": "you've ever opened the Gemini pricing",
      "source": "source_caption"
    },
    {
      "start": 902.12,
      "end": 904.24,
      "text": "page and closed it 5 minutes later still",
      "source": "source_caption"
    },
    {
      "start": 904.24,
      "end": 906.24,
      "text": "unsure which plan you need, that's not",
      "source": "source_caption"
    },
    {
      "start": 906.24,
      "end": 908.52,
      "text": "just you. Where this is actually headed,",
      "source": "source_caption"
    },
    {
      "start": 908.52,
      "end": 910.28,
      "text": "a few things are confirmed and a few are",
      "source": "source_caption"
    },
    {
      "start": 910.28,
      "end": 912.0,
      "text": "still rumor, and it's worth keeping",
      "source": "source_caption"
    },
    {
      "start": 912.0,
      "end": 914.92,
      "text": "those separate. Confirmed, Gemini 3.5",
      "source": "source_caption"
    },
    {
      "start": 914.92,
      "end": 916.76,
      "text": "Pro is currently in partner testing with",
      "source": "source_caption"
    },
    {
      "start": 916.76,
      "end": 918.839,
      "text": "a public release expected soon, and",
      "source": "source_caption"
    },
    {
      "start": 918.839,
      "end": 920.24,
      "text": "Google has already started training on",
      "source": "source_caption"
    },
    {
      "start": 920.24,
      "end": 923.44,
      "text": "Gemini 4, according to its own July 2026",
      "source": "source_caption"
    },
    {
      "start": 923.44,
      "end": 925.0,
      "text": "announcement, though there's no public",
      "source": "source_caption"
    },
    {
      "start": 925.0,
      "end": 927.16,
      "text": "timeline for that yet. Workspace AI",
      "source": "source_caption"
    },
    {
      "start": 927.16,
      "end": 929.6,
      "text": "rollout continues expanding, and Gemini",
      "source": "source_caption"
    },
    {
      "start": 929.6,
      "end": 931.28,
      "text": "Live's regional language support keeps",
      "source": "source_caption"
    },
    {
      "start": 931.28,
      "end": 933.92,
      "text": "growing. Speculative and worth labeling",
      "source": "source_caption"
    },
    {
      "start": 933.92,
      "end": 935.72,
      "text": "clearly as such, there's talk of a",
      "source": "source_caption"
    },
    {
      "start": 935.72,
      "end": 938.08,
      "text": "dedicated on-device AI chip for future",
      "source": "source_caption"
    },
    {
      "start": 938.08,
      "end": 940.04,
      "text": "Pixel phones, and some experimental",
      "source": "source_caption"
    },
    {
      "start": 940.04,
      "end": 942.48,
      "text": "DeepMind research around 3D avatars and",
      "source": "source_caption"
    },
    {
      "start": 942.48,
      "end": 944.6,
      "text": "world simulation that hasn't shipped as",
      "source": "source_caption"
    },
    {
      "start": 944.6,
      "end": 946.44,
      "text": "a product. Treat both of those as",
      "source": "source_caption"
    },
    {
      "start": 946.44,
      "end": 949.28,
      "text": "possible, not coming. Nothing official",
      "source": "source_caption"
    },
    {
      "start": 949.28,
      "end": 951.6,
      "text": "has confirmed either one. The verdict.",
      "source": "source_caption"
    },
    {
      "start": 951.6,
      "end": 953.64,
      "text": "So, where does that leave things? Gemini",
      "source": "source_caption"
    },
    {
      "start": 953.64,
      "end": 956.16,
      "text": "in 2026 isn't a chatbot you occasionally",
      "source": "source_caption"
    },
    {
      "start": 956.16,
      "end": 958.44,
      "text": "open. It's an AI layer Google has",
      "source": "source_caption"
    },
    {
      "start": 958.44,
      "end": 960.84,
      "text": "threaded through search, Gmail, your",
      "source": "source_caption"
    },
    {
      "start": 960.84,
      "end": 963.4,
      "text": "documents, and increasingly your phone",
      "source": "source_caption"
    },
    {
      "start": 963.4,
      "end": 965.88,
      "text": "itself. The models handle the thinking,",
      "source": "source_caption"
    },
    {
      "start": 965.88,
      "end": 968.2,
      "text": "the modes handle how you ask, and agents",
      "source": "source_caption"
    },
    {
      "start": 968.2,
      "end": 970.04,
      "text": "like Deep Research are the clearest sign",
      "source": "source_caption"
    },
    {
      "start": 970.04,
      "end": 971.96,
      "text": "of where all of it is actually heading.",
      "source": "source_caption"
    },
    {
      "start": 971.96,
      "end": 973.4,
      "text": "If there's one thing worth trying this",
      "source": "source_caption"
    },
    {
      "start": 973.4,
      "end": 975.52,
      "text": "week, it's Deep Research on something",
      "source": "source_caption"
    },
    {
      "start": 975.52,
      "end": 977.04,
      "text": "you'd normally spend an evening looking",
      "source": "source_caption"
    },
    {
      "start": 977.04,
      "end": 979.16,
      "text": "into yourself, and actually watching it",
      "source": "source_caption"
    },
    {
      "start": 979.16,
      "end": 980.76,
      "text": "work instead of just reading the final",
      "source": "source_caption"
    },
    {
      "start": 980.76,
      "end": 982.72,
      "text": "report. Drop a comment with which piece",
      "source": "source_caption"
    },
    {
      "start": 982.72,
      "end": 984.64,
      "text": "of this surprised you most, the model",
      "source": "source_caption"
    },
    {
      "start": 984.64,
      "end": 987.32,
      "text": "lineup, the agent side, or just how much",
      "source": "source_caption"
    },
    {
      "start": 987.32,
      "end": 988.76,
      "text": "of this you were already using without",
      "source": "source_caption"
    },
    {
      "start": 988.76,
      "end": 990.48,
      "text": "realizing it. I'll be back soon with a",
      "source": "source_caption"
    },
    {
      "start": 990.48,
      "end": 992.4,
      "text": "deeper breakdown on how Deep Research",
      "source": "source_caption"
    },
    {
      "start": 992.4,
      "end": 994.0,
      "text": "actually performs against a real",
      "source": "source_caption"
    },
    {
      "start": 994.0,
      "end": 996.08,
      "text": "research task. Thanks for watching, and",
      "source": "source_caption"
    },
    {
      "start": 996.08,
      "end": 999.2,
      "text": "I'll see you in the next one.",
      "source": "source_caption"
    }
  ],
  "transcript_selection": {
    "schema": "videonote.transcript_selection.v1",
    "status": "SELECTED",
    "requested_language": "en",
    "duration": 997.248,
    "reference_path": "/Users/sagawa/AI_WORK/video_notes/url/20260817_164301__Google_Gemini_Explained_80_of_Google_Gemini_You_Never_Use_2026_Guide/reference_subtitle.vtt",
    "reference_kind": "automatic_original",
    "reference_language": "en-orig",
    "selected_source": "youtube_auto_original",
    "decision_reason": "source_timing_retained_after_quality_gate",
    "whisper_rejected": false,
    "source_profile": {
      "segment_count": 454,
      "multilingual": false,
      "first_start": 0.0,
      "final_end": 999.2,
      "duration_coverage": 1.0,
      "maximum_gap_seconds": 0.0,
      "inverted_count": 0,
      "language": {
        "requested": "en",
        "matched": true,
        "latin": 13376,
        "han": 0,
        "kana": 0,
        "hangul": 0
      },
      "hallucination_region_count": 0,
      "hallucination_regions": [],
      "usable_full": true,
      "usable_partial": true
    },
    "whisper_profile": {
      "segment_count": 231,
      "multilingual": false,
      "first_start": 0.0,
      "final_end": 997.12,
      "duration_coverage": 0.999872,
      "maximum_gap_seconds": 1.04,
      "inverted_count": 0,
      "language": {
        "requested": "en",
        "matched": true,
        "latin": 13378,
        "han": 0,
        "kana": 0,
        "hangul": 0
      },
      "hallucination_region_count": 0,
      "hallucination_regions": [],
      "usable_full": true,
      "usable_partial": true
    },
    "comparison": {
      "window_seconds": 30.0,
      "compared_window_count": 34,
      "median_score": 0.9521,
      "divergent_window_count": 0,
      "divergent_ratio": 0.0,
      "divergent_windows": []
    },
    "timestamp_alignment": {
      "applied": false,
      "estimated_offset_seconds": 0.0,
      "matched_segment_count": 0,
      "match_scores": []
    },
    "caption_timing_repair": {
      "method": "monotonic_multi_anchor_piecewise_interpolation",
      "applied": false,
      "anchor_count": 0,
      "anchors": [],
      "source_block_count": 86,
      "whisper_block_count": 106,
      "matched_block_count": 76,
      "case_normalization": {
        "applied": false,
        "changed_segment_count": 0
      },
      "source_text_preserved": true,
      "reason": "source_timing_retained_after_quality_gate"
    },
    "filled_source_profile": {
      "segment_count": 87,
      "multilingual": false,
      "first_start": 0.0,
      "final_end": 999.2,
      "duration_coverage": 1.0,
      "maximum_gap_seconds": 0.0,
      "inverted_count": 0,
      "language": {
        "requested": "en",
        "matched": true,
        "latin": 13376,
        "han": 0,
        "kana": 0,
        "hangul": 0
      },
      "hallucination_region_count": 0,
      "hallucination_regions": [],
      "usable_full": true,
      "usable_partial": true
    },
    "whisper_duration_guard": {
      "media_duration": 997.248,
      "rejected_segment_count": 0,
      "rejected_segments": []
    }
  }
}