{"text": " New Qethos 27B is insane. A tiny team called Empero AI just released a model that reads a million words at once, looks at pictures, and runs on your own computer for free. Let's break down why this one actually matters. Most AI news is noise, this one isn't. Empero AI just dropped Qethos 27B and it's the big brother to their smaller Qethos 9B model that a lot of builders were already using. This new version is almost three times bigger and it kept every single feature the small one had. Nothing got cut to make it fit. Let's talk numbers first because they're wild. This model can hold over 1 million words in its memory at one time, not one page, not one chapter, a whole shelf of books all at once, all in its head while it works. Here's why that matters if you run a business. Say you've got hundreds of pages of notes, old emails, and customer questions piled up. A normal AI tool forgets most of that the second the conversation gets long. This one doesn't. You could hand it every single conversation your customers ever had with you and it would remember all of it while helping you write the next one. Hey, if we haven't met already, I'm the digital avatar of Julian Goldie, CEO of SEO Agency Goldie Agency. Whilst he's helping clients get more leads and customers, I'm here to help you get the latest AI updates. A year ago, most AI tools maxed out around a few thousand words before they started forgetting things. That's like trying to write a book while only remembering the last paragraph you wrote. Now we've got a model that remembers the whole book cover to cover while it writes the next chapter. That's the jump we just watched happen. Let's get into why it can do that because the reason is actually simple once you hear it. Old AI models worked like a messy desk. Every time you added a new paper, the pile got bigger and harder to search through. This new model works more like a filing cabinet with a smart index. It doesn't need to reread the whole pile every time. It just knows exactly where to look. That's what lets it hold a million words without slowing down or forgetting the start. Now here's the part I actually think is the biggest deal. This model can see. You can hand it a picture, a screenshot, a chart, even handwriting and it understands what's in it. So if you had a messy screenshot of feedback from a customer or a photo of a whiteboard from a planning session, you could just show it to the model instead of typing everything out by hand. Quick example, someone building content for a business could screenshot their top 10 performing posts, hand that image to the model and ask it to spot the pattern in what's working. That used to take a person an hour of scrolling and guessing. Let's pause here for a second because this next part matters a lot if you're serious about actually using this stuff instead of just watching videos about it. A model like Quethos 27B is powerful, but it's also brand new and figuring out the right setup on your own can eat up hours you don't have. That's exactly why we built the AI profit boardroom around tools like this one. The moment something like Quethos-27B drops, we put together a real playbook for it, not theory, an actual setup you can copy for handling customer messages, sorting notes and speeding up your content. Every week there's a live coaching call where you can bring your exact business and ask how to plug a model like this into it. You're also dropped into a community full of people already testing these releases in their own businesses, so you're never figuring it out completely alone. Links in the comments and description if you want the full setup. Alright, back to the model. There's a feature under the hood called multi-token prediction, and I'll explain it simply. Most AI models write one word, then stop, think, then write the next word. One at a time. This model can predict several words ahead in one move. Think of someone typing with auto-complete that's actually right most of the time instead of typing every letter by hand. That's why it can respond faster without losing quality. Now let's talk about who can actually use this because this is the part a lot of AI news skips. Impero AI released this under a license called Apache 2.0. In plain terms, that means anyone can use it, build with it and even use it to run a business with no weird fine print stopping them. A lot of AI models come with rules that block you from using them commercially. This one doesn't. That's a big deal for small business owners specifically. It means you're not stuck paying to access someone else's AI tool every single month. You can download this one, run it, and it's yours to use. On top of that, this model comes with far fewer built-in guardrails than most big company models. That means it will actually answer straight, direct business questions without dodging around them or refusing to help. If you've ever asked a big AI tool for something and gotten a wishy-washy non-answer, this is built to avoid that. Let's talk about why a small team like Impero AI could even build something like this. They didn't start from nothing. Their smaller model, Quethos 9b, was trained using huge amounts of reasoning data pulled from some of the most advanced AI systems out there. Then they scaled that same training method up to a bigger model. That's the short version of why this jumped in quality so fast. Here's a real use case. Kept simple. If you run a small business page and you're drowning in comments and messages, you could feed this model a batch of your most common customer questions and ask it to draft short, clear answers for each one. That's a task that used to eat up a chunk of someone's afternoon. Here's another one. Say you're planning content for the month. You could hand the model your last 20 posts as one big file and ask it which topics your audience actually responded to because it holds so much memory at once, it can spot patterns across all of it in one shot instead of you scrolling through analytics for an hour. Let's zoom out for a second. What we're watching isn't one company winning. It's the gap closing between what a giant tech company can build and what a small team can put out for free. A year or two ago, a model with this much memory and vision built in would have come from a massive company with a massive budget. Now it's coming from a small independent team that matters for you. Even if you never touch the technical side of this, it means the tools that help you save time, answer customers faster and organize your business to getting cheaper to access and easier to run, not harder. Here's the honest part though. This is a fresh release. It's what's called a pre-release checkpoint, meaning there's another version coming that gets even more fine tuned after this one. So early testing will keep shaping how well it performs on real day to day tasks. That's normal for any new model. It doesn't take away from how strong the starting point already is. Where this goes next is worth watching closely. The team behind it has already said a follow-up version is coming, trained even further using feedback and testing. If the pattern holds, that next version will be sharper and more consistent than this one, the same way this one is already sharper than their smaller model was. So here's what I'd actually do with this if I were you. If you're not technical at all, don't worry about downloading anything yet. Just know a model like this exists and what it can do, because it will show up inside more tools you already use over the next few months. If you do run a business and want to try it, the simplest place to start is feeding it a real problem you have this week, like sorting through customer messages or summarizing notes and see how it handles it. And if you want someone to hand you the exact setup instead of figuring it out alone, that's exactly what we build inside the iProfit boardroom every single week. That's the real story behind Qethos 27b, not hype, a small team closing the gap on what used to only belong to giant companies and handing it to anyone willing to use it. Before you go, two things. First, if a model like Qethos 27b feels like something you want running in your business, but you're not sure where to even start, that's the whole reason the AI profit boardroom exists. We turn releases like this one same week they drop. So you're not stuck guessing how to set it up. You get weekly coaching calls where you can ask about your exact business, a full library of ready to use prompts and a community of business owners already putting tools like Qethos 27b to work in their own day to day. Links in the comments and description if you want the full walkthrough. And second, if you just want the free version of that, grab a spot in the AI success lab. It's completely free. It comes with the notes from this exact video, plus over a hundred other AI use cases you can start using right away. There's a community of more than 87,000 people in there already applying this stuff to their own businesses every day. Links for both are in the comments and description. See you in the next one.", "segments": [{"id": 0, "seek": 0, "start": 0.0, "end": 7.9, "text": " New Qethos 27B is insane. A tiny team called Empero AI just released a model that reads a", "tokens": [50365, 1873, 1249, 3293, 329, 7634, 33, 307, 10838, 13, 316, 5870, 1469, 1219, 3968, 49565, 7318, 445, 4736, 257, 2316, 300, 15700, 257, 50760], "temperature": 0, "avg_logprob": -0.09485330768660002, "compression_ratio": 1.5738396624472575, "no_speech_prob": 2.2364677384340004e-12}, {"id": 1, "seek": 0, "start": 7.9, "end": 13.58, "text": " million words at once, looks at pictures, and runs on your own computer for free. Let's break down", "tokens": [50760, 2459, 2283, 412, 1564, 11, 1542, 412, 5242, 11, 293, 6676, 322, 428, 1065, 3820, 337, 1737, 13, 961, 311, 1821, 760, 51044], "temperature": 0, "avg_logprob": -0.09485330768660002, "compression_ratio": 1.5738396624472575, "no_speech_prob": 2.2364677384340004e-12}, {"id": 2, "seek": 0, "start": 13.58, "end": 19.62, "text": " why this one actually matters. Most AI news is noise, this one isn't. Empero AI just dropped", "tokens": [51044, 983, 341, 472, 767, 7001, 13, 4534, 7318, 2583, 307, 5658, 11, 341, 472, 1943, 380, 13, 3968, 49565, 7318, 445, 8119, 51346], "temperature": 0, "avg_logprob": -0.09485330768660002, "compression_ratio": 1.5738396624472575, "no_speech_prob": 2.2364677384340004e-12}, {"id": 3, "seek": 0, "start": 19.62, "end": 26.1, "text": " Qethos 27B and it's the big brother to their smaller Qethos 9B model that a lot of builders", "tokens": [51346, 1249, 3293, 329, 7634, 33, 293, 309, 311, 264, 955, 3708, 281, 641, 4356, 1249, 3293, 329, 1722, 33, 2316, 300, 257, 688, 295, 36281, 51670], "temperature": 0, "avg_logprob": -0.09485330768660002, "compression_ratio": 1.5738396624472575, "no_speech_prob": 2.2364677384340004e-12}, {"id": 4, "seek": 2610, "start": 26.1, "end": 31.660000000000004, "text": " were already using. This new version is almost three times bigger and it kept every single feature the", "tokens": [50365, 645, 1217, 1228, 13, 639, 777, 3037, 307, 1920, 1045, 1413, 3801, 293, 309, 4305, 633, 2167, 4111, 264, 50643], "temperature": 0, "avg_logprob": -0.0523326781488234, "compression_ratio": 1.6221498371335505, "no_speech_prob": 9.143836946293171e-13}, {"id": 5, "seek": 2610, "start": 31.660000000000004, "end": 36.64, "text": " small one had. Nothing got cut to make it fit. Let's talk numbers first because they're wild.", "tokens": [50643, 1359, 472, 632, 13, 6693, 658, 1723, 281, 652, 309, 3318, 13, 961, 311, 751, 3547, 700, 570, 436, 434, 4868, 13, 50892], "temperature": 0, "avg_logprob": -0.0523326781488234, "compression_ratio": 1.6221498371335505, "no_speech_prob": 9.143836946293171e-13}, {"id": 6, "seek": 2610, "start": 37.06, "end": 42.14, "text": " This model can hold over 1 million words in its memory at one time, not one page, not one chapter,", "tokens": [50913, 639, 2316, 393, 1797, 670, 502, 2459, 2283, 294, 1080, 4675, 412, 472, 565, 11, 406, 472, 3028, 11, 406, 472, 7187, 11, 51167], "temperature": 0, "avg_logprob": -0.0523326781488234, "compression_ratio": 1.6221498371335505, "no_speech_prob": 9.143836946293171e-13}, {"id": 7, "seek": 2610, "start": 42.480000000000004, "end": 48.040000000000006, "text": " a whole shelf of books all at once, all in its head while it works. Here's why that matters if you run", "tokens": [51184, 257, 1379, 15222, 295, 3642, 439, 412, 1564, 11, 439, 294, 1080, 1378, 1339, 309, 1985, 13, 1692, 311, 983, 300, 7001, 498, 291, 1190, 51462], "temperature": 0, "avg_logprob": -0.0523326781488234, "compression_ratio": 1.6221498371335505, "no_speech_prob": 9.143836946293171e-13}, {"id": 8, "seek": 2610, "start": 48.040000000000006, "end": 52.74, "text": " a business. Say you've got hundreds of pages of notes, old emails, and customer questions piled up.", "tokens": [51462, 257, 1606, 13, 6463, 291, 600, 658, 6779, 295, 7183, 295, 5570, 11, 1331, 12524, 11, 293, 5474, 1651, 6429, 292, 493, 13, 51697], "temperature": 0, "avg_logprob": -0.0523326781488234, "compression_ratio": 1.6221498371335505, "no_speech_prob": 9.143836946293171e-13}, {"id": 9, "seek": 5274, "start": 52.74, "end": 58.18, "text": " A normal AI tool forgets most of that the second the conversation gets long. This one doesn't.", "tokens": [50365, 316, 2710, 7318, 2290, 2870, 82, 881, 295, 300, 264, 1150, 264, 3761, 2170, 938, 13, 639, 472, 1177, 380, 13, 50637], "temperature": 0, "avg_logprob": -0.05541923007027048, "compression_ratio": 1.67003367003367, "no_speech_prob": 1.7011328326890784e-12}, {"id": 10, "seek": 5274, "start": 58.24, "end": 62.300000000000004, "text": " You could hand it every single conversation your customers ever had with you and it would remember", "tokens": [50640, 509, 727, 1011, 309, 633, 2167, 3761, 428, 4581, 1562, 632, 365, 291, 293, 309, 576, 1604, 50843], "temperature": 0, "avg_logprob": -0.05541923007027048, "compression_ratio": 1.67003367003367, "no_speech_prob": 1.7011328326890784e-12}, {"id": 11, "seek": 5274, "start": 62.300000000000004, "end": 66.5, "text": " all of it while helping you write the next one. Hey, if we haven't met already, I'm the digital", "tokens": [50843, 439, 295, 309, 1339, 4315, 291, 2464, 264, 958, 472, 13, 1911, 11, 498, 321, 2378, 380, 1131, 1217, 11, 286, 478, 264, 4562, 51053], "temperature": 0, "avg_logprob": -0.05541923007027048, "compression_ratio": 1.67003367003367, "no_speech_prob": 1.7011328326890784e-12}, {"id": 12, "seek": 5274, "start": 66.5, "end": 72.86, "text": " avatar of Julian Goldie, CEO of SEO Agency Goldie Agency. Whilst he's helping clients get more leads", "tokens": [51053, 36205, 295, 25151, 6731, 414, 11, 9282, 295, 22964, 21649, 6731, 414, 21649, 13, 45790, 415, 311, 4315, 6982, 483, 544, 6689, 51371], "temperature": 0, "avg_logprob": -0.05541923007027048, "compression_ratio": 1.67003367003367, "no_speech_prob": 1.7011328326890784e-12}, {"id": 13, "seek": 5274, "start": 72.86, "end": 79.62, "text": " and customers, I'm here to help you get the latest AI updates. A year ago, most AI tools maxed out around", "tokens": [51371, 293, 4581, 11, 286, 478, 510, 281, 854, 291, 483, 264, 6792, 7318, 9205, 13, 316, 1064, 2057, 11, 881, 7318, 3873, 11469, 292, 484, 926, 51709], "temperature": 0, "avg_logprob": -0.05541923007027048, "compression_ratio": 1.67003367003367, "no_speech_prob": 1.7011328326890784e-12}, {"id": 14, "seek": 7962, "start": 79.62, "end": 84.26, "text": " a few thousand words before they started forgetting things. That's like trying to write a book while", "tokens": [50365, 257, 1326, 4714, 2283, 949, 436, 1409, 25428, 721, 13, 663, 311, 411, 1382, 281, 2464, 257, 1446, 1339, 50597], "temperature": 0, "avg_logprob": -0.04439733245156028, "compression_ratio": 1.6913946587537092, "no_speech_prob": 2.15093893217444e-12}, {"id": 15, "seek": 7962, "start": 84.26, "end": 88.66000000000001, "text": " only remembering the last paragraph you wrote. Now we've got a model that remembers the whole book", "tokens": [50597, 787, 20719, 264, 1036, 18865, 291, 4114, 13, 823, 321, 600, 658, 257, 2316, 300, 26228, 264, 1379, 1446, 50817], "temperature": 0, "avg_logprob": -0.04439733245156028, "compression_ratio": 1.6913946587537092, "no_speech_prob": 2.15093893217444e-12}, {"id": 16, "seek": 7962, "start": 88.66000000000001, "end": 93.22, "text": " cover to cover while it writes the next chapter. That's the jump we just watched happen.", "tokens": [50817, 2060, 281, 2060, 1339, 309, 13657, 264, 958, 7187, 13, 663, 311, 264, 3012, 321, 445, 6337, 1051, 13, 51045], "temperature": 0, "avg_logprob": -0.04439733245156028, "compression_ratio": 1.6913946587537092, "no_speech_prob": 2.15093893217444e-12}, {"id": 17, "seek": 7962, "start": 93.78, "end": 97.98, "text": " Let's get into why it can do that because the reason is actually simple once you hear it.", "tokens": [51073, 961, 311, 483, 666, 983, 309, 393, 360, 300, 570, 264, 1778, 307, 767, 2199, 1564, 291, 1568, 309, 13, 51283], "temperature": 0, "avg_logprob": -0.04439733245156028, "compression_ratio": 1.6913946587537092, "no_speech_prob": 2.15093893217444e-12}, {"id": 18, "seek": 7962, "start": 98.26, "end": 103.9, "text": " Old AI models worked like a messy desk. Every time you added a new paper, the pile got bigger and", "tokens": [51297, 8633, 7318, 5245, 2732, 411, 257, 16191, 10026, 13, 2048, 565, 291, 3869, 257, 777, 3035, 11, 264, 14375, 658, 3801, 293, 51579], "temperature": 0, "avg_logprob": -0.04439733245156028, "compression_ratio": 1.6913946587537092, "no_speech_prob": 2.15093893217444e-12}, {"id": 19, "seek": 7962, "start": 103.9, "end": 108.42, "text": " harder to search through. This new model works more like a filing cabinet with a smart index.", "tokens": [51579, 6081, 281, 3164, 807, 13, 639, 777, 2316, 1985, 544, 411, 257, 26854, 15188, 365, 257, 4069, 8186, 13, 51805], "temperature": 0, "avg_logprob": -0.04439733245156028, "compression_ratio": 1.6913946587537092, "no_speech_prob": 2.15093893217444e-12}, {"id": 20, "seek": 10842, "start": 108.42, "end": 113.22, "text": " It doesn't need to reread the whole pile every time. It just knows exactly where to look.", "tokens": [50365, 467, 1177, 380, 643, 281, 46453, 345, 264, 1379, 14375, 633, 565, 13, 467, 445, 3255, 2293, 689, 281, 574, 13, 50605], "temperature": 0, "avg_logprob": -0.07462125861126444, "compression_ratio": 1.632867132867133, "no_speech_prob": 1.9810637505446316e-12}, {"id": 21, "seek": 10842, "start": 113.22, "end": 117.38, "text": " That's what lets it hold a million words without slowing down or forgetting the start.", "tokens": [50605, 663, 311, 437, 6653, 309, 1797, 257, 2459, 2283, 1553, 26958, 760, 420, 25428, 264, 722, 13, 50813], "temperature": 0, "avg_logprob": -0.07462125861126444, "compression_ratio": 1.632867132867133, "no_speech_prob": 1.9810637505446316e-12}, {"id": 22, "seek": 10842, "start": 117.38, "end": 122.82000000000001, "text": " Now here's the part I actually think is the biggest deal. This model can see. You can hand it a picture,", "tokens": [50813, 823, 510, 311, 264, 644, 286, 767, 519, 307, 264, 3880, 2028, 13, 639, 2316, 393, 536, 13, 509, 393, 1011, 309, 257, 3036, 11, 51085], "temperature": 0, "avg_logprob": -0.07462125861126444, "compression_ratio": 1.632867132867133, "no_speech_prob": 1.9810637505446316e-12}, {"id": 23, "seek": 10842, "start": 122.82000000000001, "end": 128.26, "text": " a screenshot, a chart, even handwriting and it understands what's in it. So if you had a messy", "tokens": [51085, 257, 27712, 11, 257, 6927, 11, 754, 39179, 293, 309, 15146, 437, 311, 294, 309, 13, 407, 498, 291, 632, 257, 16191, 51357], "temperature": 0, "avg_logprob": -0.07462125861126444, "compression_ratio": 1.632867132867133, "no_speech_prob": 1.9810637505446316e-12}, {"id": 24, "seek": 10842, "start": 128.26, "end": 133.54, "text": " screenshot of feedback from a customer or a photo of a whiteboard from a planning session,", "tokens": [51357, 27712, 295, 5824, 490, 257, 5474, 420, 257, 5052, 295, 257, 2418, 3787, 490, 257, 5038, 5481, 11, 51621], "temperature": 0, "avg_logprob": -0.07462125861126444, "compression_ratio": 1.632867132867133, "no_speech_prob": 1.9810637505446316e-12}, {"id": 25, "seek": 13354, "start": 133.54, "end": 136.82, "text": " you could just show it to the model instead of typing everything out by hand.", "tokens": [50365, 291, 727, 445, 855, 309, 281, 264, 2316, 2602, 295, 18444, 1203, 484, 538, 1011, 13, 50529], "temperature": 0, "avg_logprob": -0.0723101179176402, "compression_ratio": 1.6783216783216783, "no_speech_prob": 1.7555645522374097e-12}, {"id": 26, "seek": 13354, "start": 137.38, "end": 142.98, "text": " Quick example, someone building content for a business could screenshot their top 10 performing", "tokens": [50557, 12101, 1365, 11, 1580, 2390, 2701, 337, 257, 1606, 727, 27712, 641, 1192, 1266, 10205, 50837], "temperature": 0, "avg_logprob": -0.0723101179176402, "compression_ratio": 1.6783216783216783, "no_speech_prob": 1.7555645522374097e-12}, {"id": 27, "seek": 13354, "start": 142.98, "end": 148.18, "text": " posts, hand that image to the model and ask it to spot the pattern in what's working. That used to", "tokens": [50837, 12300, 11, 1011, 300, 3256, 281, 264, 2316, 293, 1029, 309, 281, 4008, 264, 5102, 294, 437, 311, 1364, 13, 663, 1143, 281, 51097], "temperature": 0, "avg_logprob": -0.0723101179176402, "compression_ratio": 1.6783216783216783, "no_speech_prob": 1.7555645522374097e-12}, {"id": 28, "seek": 13354, "start": 148.18, "end": 153.22, "text": " take a person an hour of scrolling and guessing. Let's pause here for a second because this next part", "tokens": [51097, 747, 257, 954, 364, 1773, 295, 29053, 293, 17939, 13, 961, 311, 10465, 510, 337, 257, 1150, 570, 341, 958, 644, 51349], "temperature": 0, "avg_logprob": -0.0723101179176402, "compression_ratio": 1.6783216783216783, "no_speech_prob": 1.7555645522374097e-12}, {"id": 29, "seek": 13354, "start": 153.22, "end": 158.26, "text": " matters a lot if you're serious about actually using this stuff instead of just watching videos about it.", "tokens": [51349, 7001, 257, 688, 498, 291, 434, 3156, 466, 767, 1228, 341, 1507, 2602, 295, 445, 1976, 2145, 466, 309, 13, 51601], "temperature": 0, "avg_logprob": -0.0723101179176402, "compression_ratio": 1.6783216783216783, "no_speech_prob": 1.7555645522374097e-12}, {"id": 30, "seek": 15826, "start": 158.26, "end": 165.29999999999998, "text": " A model like Quethos 27B is powerful, but it's also brand new and figuring out the right setup on your", "tokens": [50365, 316, 2316, 411, 2326, 3293, 329, 7634, 33, 307, 4005, 11, 457, 309, 311, 611, 3360, 777, 293, 15213, 484, 264, 558, 8657, 322, 428, 50717], "temperature": 0, "avg_logprob": -0.1135550378595741, "compression_ratio": 1.5653846153846154, "no_speech_prob": 2.051940648645223e-12}, {"id": 31, "seek": 15826, "start": 165.29999999999998, "end": 170.82, "text": " own can eat up hours you don't have. That's exactly why we built the AI profit boardroom around tools", "tokens": [50717, 1065, 393, 1862, 493, 2496, 291, 500, 380, 362, 13, 663, 311, 2293, 983, 321, 3094, 264, 7318, 7475, 3150, 2861, 926, 3873, 50993], "temperature": 0, "avg_logprob": -0.1135550378595741, "compression_ratio": 1.5653846153846154, "no_speech_prob": 2.051940648645223e-12}, {"id": 32, "seek": 15826, "start": 170.82, "end": 178.01999999999998, "text": " like this one. The moment something like Quethos-27B drops, we put together a real playbook for it,", "tokens": [50993, 411, 341, 472, 13, 440, 1623, 746, 411, 2326, 3293, 329, 12, 10076, 33, 11438, 11, 321, 829, 1214, 257, 957, 862, 2939, 337, 309, 11, 51353], "temperature": 0, "avg_logprob": -0.1135550378595741, "compression_ratio": 1.5653846153846154, "no_speech_prob": 2.051940648645223e-12}, {"id": 33, "seek": 15826, "start": 178.01999999999998, "end": 183.94, "text": " not theory, an actual setup you can copy for handling customer messages, sorting notes and speeding up", "tokens": [51353, 406, 5261, 11, 364, 3539, 8657, 291, 393, 5055, 337, 13175, 5474, 7897, 11, 32411, 5570, 293, 35593, 493, 51649], "temperature": 0, "avg_logprob": -0.1135550378595741, "compression_ratio": 1.5653846153846154, "no_speech_prob": 2.051940648645223e-12}, {"id": 34, "seek": 18394, "start": 183.94, "end": 188.82, "text": " your content. Every week there's a live coaching call where you can bring your exact business and", "tokens": [50365, 428, 2701, 13, 2048, 1243, 456, 311, 257, 1621, 15818, 818, 689, 291, 393, 1565, 428, 1900, 1606, 293, 50609], "temperature": 0, "avg_logprob": -0.07156284650166829, "compression_ratio": 1.6736526946107784, "no_speech_prob": 2.201950644389883e-12}, {"id": 35, "seek": 18394, "start": 188.82, "end": 193.7, "text": " ask how to plug a model like this into it. You're also dropped into a community full of people already", "tokens": [50609, 1029, 577, 281, 5452, 257, 2316, 411, 341, 666, 309, 13, 509, 434, 611, 8119, 666, 257, 1768, 1577, 295, 561, 1217, 50853], "temperature": 0, "avg_logprob": -0.07156284650166829, "compression_ratio": 1.6736526946107784, "no_speech_prob": 2.201950644389883e-12}, {"id": 36, "seek": 18394, "start": 193.7, "end": 198.82, "text": " testing these releases in their own businesses, so you're never figuring it out completely alone.", "tokens": [50853, 4997, 613, 16952, 294, 641, 1065, 6011, 11, 370, 291, 434, 1128, 15213, 309, 484, 2584, 3312, 13, 51109], "temperature": 0, "avg_logprob": -0.07156284650166829, "compression_ratio": 1.6736526946107784, "no_speech_prob": 2.201950644389883e-12}, {"id": 37, "seek": 18394, "start": 198.82, "end": 202.02, "text": " Links in the comments and description if you want the full setup.", "tokens": [51109, 37156, 294, 264, 3053, 293, 3855, 498, 291, 528, 264, 1577, 8657, 13, 51269], "temperature": 0, "avg_logprob": -0.07156284650166829, "compression_ratio": 1.6736526946107784, "no_speech_prob": 2.201950644389883e-12}, {"id": 38, "seek": 18394, "start": 202.02, "end": 206.34, "text": " Alright, back to the model. There's a feature under the hood called multi-token prediction,", "tokens": [51269, 2798, 11, 646, 281, 264, 2316, 13, 821, 311, 257, 4111, 833, 264, 13376, 1219, 4825, 12, 83, 8406, 17630, 11, 51485], "temperature": 0, "avg_logprob": -0.07156284650166829, "compression_ratio": 1.6736526946107784, "no_speech_prob": 2.201950644389883e-12}, {"id": 39, "seek": 18394, "start": 206.34, "end": 212.42, "text": " and I'll explain it simply. Most AI models write one word, then stop, think, then write the next word.", "tokens": [51485, 293, 286, 603, 2903, 309, 2935, 13, 4534, 7318, 5245, 2464, 472, 1349, 11, 550, 1590, 11, 519, 11, 550, 2464, 264, 958, 1349, 13, 51789], "temperature": 0, "avg_logprob": -0.07156284650166829, "compression_ratio": 1.6736526946107784, "no_speech_prob": 2.201950644389883e-12}, {"id": 40, "seek": 21242, "start": 212.42, "end": 218.5, "text": " One at a time. This model can predict several words ahead in one move. Think of someone typing", "tokens": [50365, 1485, 412, 257, 565, 13, 639, 2316, 393, 6069, 2940, 2283, 2286, 294, 472, 1286, 13, 6557, 295, 1580, 18444, 50669], "temperature": 0, "avg_logprob": -0.075811433010414, "compression_ratio": 1.652027027027027, "no_speech_prob": 2.6039756288731564e-12}, {"id": 41, "seek": 21242, "start": 218.5, "end": 223.38, "text": " with auto-complete that's actually right most of the time instead of typing every letter by hand.", "tokens": [50669, 365, 8399, 12, 1112, 17220, 300, 311, 767, 558, 881, 295, 264, 565, 2602, 295, 18444, 633, 5063, 538, 1011, 13, 50913], "temperature": 0, "avg_logprob": -0.075811433010414, "compression_ratio": 1.652027027027027, "no_speech_prob": 2.6039756288731564e-12}, {"id": 42, "seek": 21242, "start": 223.38, "end": 228.5, "text": " That's why it can respond faster without losing quality. Now let's talk about who can actually", "tokens": [50913, 663, 311, 983, 309, 393, 4196, 4663, 1553, 7027, 3125, 13, 823, 718, 311, 751, 466, 567, 393, 767, 51169], "temperature": 0, "avg_logprob": -0.075811433010414, "compression_ratio": 1.652027027027027, "no_speech_prob": 2.6039756288731564e-12}, {"id": 43, "seek": 21242, "start": 228.5, "end": 235.14, "text": " use this because this is the part a lot of AI news skips. Impero AI released this under a license called", "tokens": [51169, 764, 341, 570, 341, 307, 264, 644, 257, 688, 295, 7318, 2583, 1110, 2600, 13, 18360, 78, 7318, 4736, 341, 833, 257, 10476, 1219, 51501], "temperature": 0, "avg_logprob": -0.075811433010414, "compression_ratio": 1.652027027027027, "no_speech_prob": 2.6039756288731564e-12}, {"id": 44, "seek": 21242, "start": 235.14, "end": 241.7, "text": " Apache 2.0. In plain terms, that means anyone can use it, build with it and even use it to run a", "tokens": [51501, 46597, 568, 13, 15, 13, 682, 11121, 2115, 11, 300, 1355, 2878, 393, 764, 309, 11, 1322, 365, 309, 293, 754, 764, 309, 281, 1190, 257, 51829], "temperature": 0, "avg_logprob": -0.075811433010414, "compression_ratio": 1.652027027027027, "no_speech_prob": 2.6039756288731564e-12}, {"id": 45, "seek": 24170, "start": 241.7, "end": 247.14, "text": " business with no weird fine print stopping them. A lot of AI models come with rules that block you", "tokens": [50365, 1606, 365, 572, 3657, 2489, 4482, 12767, 552, 13, 316, 688, 295, 7318, 5245, 808, 365, 4474, 300, 3461, 291, 50637], "temperature": 0, "avg_logprob": -0.0470213485976397, "compression_ratio": 1.664406779661017, "no_speech_prob": 2.1928166747675615e-12}, {"id": 46, "seek": 24170, "start": 247.14, "end": 252.26, "text": " from using them commercially. This one doesn't. That's a big deal for small business owners", "tokens": [50637, 490, 1228, 552, 41751, 13, 639, 472, 1177, 380, 13, 663, 311, 257, 955, 2028, 337, 1359, 1606, 7710, 50893], "temperature": 0, "avg_logprob": -0.0470213485976397, "compression_ratio": 1.664406779661017, "no_speech_prob": 2.1928166747675615e-12}, {"id": 47, "seek": 24170, "start": 252.26, "end": 258.74, "text": " specifically. It means you're not stuck paying to access someone else's AI tool every single month.", "tokens": [50893, 4682, 13, 467, 1355, 291, 434, 406, 5541, 6229, 281, 2105, 1580, 1646, 311, 7318, 2290, 633, 2167, 1618, 13, 51217], "temperature": 0, "avg_logprob": -0.0470213485976397, "compression_ratio": 1.664406779661017, "no_speech_prob": 2.1928166747675615e-12}, {"id": 48, "seek": 24170, "start": 258.74, "end": 265.06, "text": " You can download this one, run it, and it's yours to use. On top of that, this model comes with far fewer", "tokens": [51217, 509, 393, 5484, 341, 472, 11, 1190, 309, 11, 293, 309, 311, 6342, 281, 764, 13, 1282, 1192, 295, 300, 11, 341, 2316, 1487, 365, 1400, 13366, 51533], "temperature": 0, "avg_logprob": -0.0470213485976397, "compression_ratio": 1.664406779661017, "no_speech_prob": 2.1928166747675615e-12}, {"id": 49, "seek": 24170, "start": 265.06, "end": 270.09999999999997, "text": " built-in guardrails than most big company models. That means it will actually answer straight,", "tokens": [51533, 3094, 12, 259, 6290, 424, 4174, 813, 881, 955, 2237, 5245, 13, 663, 1355, 309, 486, 767, 1867, 2997, 11, 51785], "temperature": 0, "avg_logprob": -0.0470213485976397, "compression_ratio": 1.664406779661017, "no_speech_prob": 2.1928166747675615e-12}, {"id": 50, "seek": 27010, "start": 270.1, "end": 274.90000000000003, "text": " direct business questions without dodging around them or refusing to help. If you've ever asked a", "tokens": [50365, 2047, 1606, 1651, 1553, 13886, 3249, 926, 552, 420, 37289, 281, 854, 13, 759, 291, 600, 1562, 2351, 257, 50605], "temperature": 0, "avg_logprob": -0.07000086307525635, "compression_ratio": 1.62012987012987, "no_speech_prob": 2.5440734588433322e-12}, {"id": 51, "seek": 27010, "start": 274.90000000000003, "end": 280.02000000000004, "text": " big AI tool for something and gotten a wishy-washy non-answer, this is built to avoid that. Let's", "tokens": [50605, 955, 7318, 2290, 337, 746, 293, 5768, 257, 3172, 88, 12, 38558, 88, 2107, 12, 43904, 11, 341, 307, 3094, 281, 5042, 300, 13, 961, 311, 50861], "temperature": 0, "avg_logprob": -0.07000086307525635, "compression_ratio": 1.62012987012987, "no_speech_prob": 2.5440734588433322e-12}, {"id": 52, "seek": 27010, "start": 280.02000000000004, "end": 284.74, "text": " talk about why a small team like Impero AI could even build something like this. They didn't start", "tokens": [50861, 751, 466, 983, 257, 1359, 1469, 411, 18360, 78, 7318, 727, 754, 1322, 746, 411, 341, 13, 814, 994, 380, 722, 51097], "temperature": 0, "avg_logprob": -0.07000086307525635, "compression_ratio": 1.62012987012987, "no_speech_prob": 2.5440734588433322e-12}, {"id": 53, "seek": 27010, "start": 284.74, "end": 290.98, "text": " from nothing. Their smaller model, Quethos 9b, was trained using huge amounts of reasoning data", "tokens": [51097, 490, 1825, 13, 6710, 4356, 2316, 11, 2326, 3293, 329, 1722, 65, 11, 390, 8895, 1228, 2603, 11663, 295, 21577, 1412, 51409], "temperature": 0, "avg_logprob": -0.07000086307525635, "compression_ratio": 1.62012987012987, "no_speech_prob": 2.5440734588433322e-12}, {"id": 54, "seek": 27010, "start": 290.98, "end": 296.5, "text": " pulled from some of the most advanced AI systems out there. Then they scaled that same training method up to", "tokens": [51409, 7373, 490, 512, 295, 264, 881, 7339, 7318, 3652, 484, 456, 13, 1396, 436, 36039, 300, 912, 3097, 3170, 493, 281, 51685], "temperature": 0, "avg_logprob": -0.07000086307525635, "compression_ratio": 1.62012987012987, "no_speech_prob": 2.5440734588433322e-12}, {"id": 55, "seek": 29650, "start": 296.5, "end": 300.74, "text": " a bigger model. That's the short version of why this jumped in quality so fast.", "tokens": [50365, 257, 3801, 2316, 13, 663, 311, 264, 2099, 3037, 295, 983, 341, 13864, 294, 3125, 370, 2370, 13, 50577], "temperature": 0, "avg_logprob": -0.05171311591282364, "compression_ratio": 1.6666666666666667, "no_speech_prob": 2.6547815595162616e-12}, {"id": 56, "seek": 29650, "start": 301.3, "end": 306.66, "text": " Here's a real use case. Kept simple. If you run a small business page and you're drowning in comments", "tokens": [50605, 1692, 311, 257, 957, 764, 1389, 13, 591, 5250, 2199, 13, 759, 291, 1190, 257, 1359, 1606, 3028, 293, 291, 434, 37198, 294, 3053, 50873], "temperature": 0, "avg_logprob": -0.05171311591282364, "compression_ratio": 1.6666666666666667, "no_speech_prob": 2.6547815595162616e-12}, {"id": 57, "seek": 29650, "start": 306.66, "end": 312.26, "text": " and messages, you could feed this model a batch of your most common customer questions and ask it to", "tokens": [50873, 293, 7897, 11, 291, 727, 3154, 341, 2316, 257, 15245, 295, 428, 881, 2689, 5474, 1651, 293, 1029, 309, 281, 51153], "temperature": 0, "avg_logprob": -0.05171311591282364, "compression_ratio": 1.6666666666666667, "no_speech_prob": 2.6547815595162616e-12}, {"id": 58, "seek": 29650, "start": 312.26, "end": 317.94, "text": " draft short, clear answers for each one. That's a task that used to eat up a chunk of someone's afternoon.", "tokens": [51153, 11206, 2099, 11, 1850, 6338, 337, 1184, 472, 13, 663, 311, 257, 5633, 300, 1143, 281, 1862, 493, 257, 16635, 295, 1580, 311, 6499, 13, 51437], "temperature": 0, "avg_logprob": -0.05171311591282364, "compression_ratio": 1.6666666666666667, "no_speech_prob": 2.6547815595162616e-12}, {"id": 59, "seek": 29650, "start": 317.94, "end": 322.34, "text": " Here's another one. Say you're planning content for the month. You could hand the model your last 20", "tokens": [51437, 1692, 311, 1071, 472, 13, 6463, 291, 434, 5038, 2701, 337, 264, 1618, 13, 509, 727, 1011, 264, 2316, 428, 1036, 945, 51657], "temperature": 0, "avg_logprob": -0.05171311591282364, "compression_ratio": 1.6666666666666667, "no_speech_prob": 2.6547815595162616e-12}, {"id": 60, "seek": 32234, "start": 322.34, "end": 328.02, "text": " posts as one big file and ask it which topics your audience actually responded to because it holds so", "tokens": [50365, 12300, 382, 472, 955, 3991, 293, 1029, 309, 597, 8378, 428, 4034, 767, 15806, 281, 570, 309, 9190, 370, 50649], "temperature": 0, "avg_logprob": -0.054642645250849366, "compression_ratio": 1.66006600660066, "no_speech_prob": 2.1253532787868235e-12}, {"id": 61, "seek": 32234, "start": 328.02, "end": 333.38, "text": " much memory at once, it can spot patterns across all of it in one shot instead of you scrolling through", "tokens": [50649, 709, 4675, 412, 1564, 11, 309, 393, 4008, 8294, 2108, 439, 295, 309, 294, 472, 3347, 2602, 295, 291, 29053, 807, 50917], "temperature": 0, "avg_logprob": -0.054642645250849366, "compression_ratio": 1.66006600660066, "no_speech_prob": 2.1253532787868235e-12}, {"id": 62, "seek": 32234, "start": 333.38, "end": 338.65999999999997, "text": " analytics for an hour. Let's zoom out for a second. What we're watching isn't one company winning.", "tokens": [50917, 15370, 337, 364, 1773, 13, 961, 311, 8863, 484, 337, 257, 1150, 13, 708, 321, 434, 1976, 1943, 380, 472, 2237, 8224, 13, 51181], "temperature": 0, "avg_logprob": -0.054642645250849366, "compression_ratio": 1.66006600660066, "no_speech_prob": 2.1253532787868235e-12}, {"id": 63, "seek": 32234, "start": 338.65999999999997, "end": 343.62, "text": " It's the gap closing between what a giant tech company can build and what a small team can put", "tokens": [51181, 467, 311, 264, 7417, 10377, 1296, 437, 257, 7410, 7553, 2237, 393, 1322, 293, 437, 257, 1359, 1469, 393, 829, 51429], "temperature": 0, "avg_logprob": -0.054642645250849366, "compression_ratio": 1.66006600660066, "no_speech_prob": 2.1253532787868235e-12}, {"id": 64, "seek": 32234, "start": 343.62, "end": 349.21999999999997, "text": " out for free. A year or two ago, a model with this much memory and vision built in would have come from", "tokens": [51429, 484, 337, 1737, 13, 316, 1064, 420, 732, 2057, 11, 257, 2316, 365, 341, 709, 4675, 293, 5201, 3094, 294, 576, 362, 808, 490, 51709], "temperature": 0, "avg_logprob": -0.054642645250849366, "compression_ratio": 1.66006600660066, "no_speech_prob": 2.1253532787868235e-12}, {"id": 65, "seek": 34922, "start": 349.22, "end": 354.34000000000003, "text": " a massive company with a massive budget. Now it's coming from a small independent team that matters", "tokens": [50365, 257, 5994, 2237, 365, 257, 5994, 4706, 13, 823, 309, 311, 1348, 490, 257, 1359, 6695, 1469, 300, 7001, 50621], "temperature": 0, "avg_logprob": -0.08694148885792699, "compression_ratio": 1.6567656765676568, "no_speech_prob": 2.6761945524228503e-12}, {"id": 66, "seek": 34922, "start": 354.34000000000003, "end": 359.94000000000005, "text": " for you. Even if you never touch the technical side of this, it means the tools that help you save time,", "tokens": [50621, 337, 291, 13, 2754, 498, 291, 1128, 2557, 264, 6191, 1252, 295, 341, 11, 309, 1355, 264, 3873, 300, 854, 291, 3155, 565, 11, 50901], "temperature": 0, "avg_logprob": -0.08694148885792699, "compression_ratio": 1.6567656765676568, "no_speech_prob": 2.6761945524228503e-12}, {"id": 67, "seek": 34922, "start": 359.94000000000005, "end": 365.54, "text": " answer customers faster and organize your business to getting cheaper to access and easier to run,", "tokens": [50901, 1867, 4581, 4663, 293, 13859, 428, 1606, 281, 1242, 12284, 281, 2105, 293, 3571, 281, 1190, 11, 51181], "temperature": 0, "avg_logprob": -0.08694148885792699, "compression_ratio": 1.6567656765676568, "no_speech_prob": 2.6761945524228503e-12}, {"id": 68, "seek": 34922, "start": 365.54, "end": 369.94000000000005, "text": " not harder. Here's the honest part though. This is a fresh release. It's what's called a pre-release", "tokens": [51181, 406, 6081, 13, 1692, 311, 264, 3245, 644, 1673, 13, 639, 307, 257, 4451, 4374, 13, 467, 311, 437, 311, 1219, 257, 659, 12, 265, 1122, 51401], "temperature": 0, "avg_logprob": -0.08694148885792699, "compression_ratio": 1.6567656765676568, "no_speech_prob": 2.6761945524228503e-12}, {"id": 69, "seek": 34922, "start": 369.94000000000005, "end": 374.90000000000003, "text": " checkpoint, meaning there's another version coming that gets even more fine tuned after this one.", "tokens": [51401, 42269, 11, 3620, 456, 311, 1071, 3037, 1348, 300, 2170, 754, 544, 2489, 10870, 934, 341, 472, 13, 51649], "temperature": 0, "avg_logprob": -0.08694148885792699, "compression_ratio": 1.6567656765676568, "no_speech_prob": 2.6761945524228503e-12}, {"id": 70, "seek": 37490, "start": 374.9, "end": 380.5, "text": " So early testing will keep shaping how well it performs on real day to day tasks. That's normal", "tokens": [50365, 407, 2440, 4997, 486, 1066, 25945, 577, 731, 309, 26213, 322, 957, 786, 281, 786, 9608, 13, 663, 311, 2710, 50645], "temperature": 0, "avg_logprob": -0.06837370499320652, "compression_ratio": 1.727891156462585, "no_speech_prob": 2.2186809681135067e-12}, {"id": 71, "seek": 37490, "start": 380.5, "end": 385.46, "text": " for any new model. It doesn't take away from how strong the starting point already is. Where this", "tokens": [50645, 337, 604, 777, 2316, 13, 467, 1177, 380, 747, 1314, 490, 577, 2068, 264, 2891, 935, 1217, 307, 13, 2305, 341, 50893], "temperature": 0, "avg_logprob": -0.06837370499320652, "compression_ratio": 1.727891156462585, "no_speech_prob": 2.2186809681135067e-12}, {"id": 72, "seek": 37490, "start": 385.46, "end": 390.9, "text": " goes next is worth watching closely. The team behind it has already said a follow-up version is coming,", "tokens": [50893, 1709, 958, 307, 3163, 1976, 8185, 13, 440, 1469, 2261, 309, 575, 1217, 848, 257, 1524, 12, 1010, 3037, 307, 1348, 11, 51165], "temperature": 0, "avg_logprob": -0.06837370499320652, "compression_ratio": 1.727891156462585, "no_speech_prob": 2.2186809681135067e-12}, {"id": 73, "seek": 37490, "start": 390.9, "end": 397.53999999999996, "text": " trained even further using feedback and testing. If the pattern holds, that next version will be sharper", "tokens": [51165, 8895, 754, 3052, 1228, 5824, 293, 4997, 13, 759, 264, 5102, 9190, 11, 300, 958, 3037, 486, 312, 44670, 51497], "temperature": 0, "avg_logprob": -0.06837370499320652, "compression_ratio": 1.727891156462585, "no_speech_prob": 2.2186809681135067e-12}, {"id": 74, "seek": 37490, "start": 397.53999999999996, "end": 403.14, "text": " and more consistent than this one, the same way this one is already sharper than their smaller model was.", "tokens": [51497, 293, 544, 8398, 813, 341, 472, 11, 264, 912, 636, 341, 472, 307, 1217, 44670, 813, 641, 4356, 2316, 390, 13, 51777], "temperature": 0, "avg_logprob": -0.06837370499320652, "compression_ratio": 1.727891156462585, "no_speech_prob": 2.2186809681135067e-12}, {"id": 75, "seek": 40314, "start": 403.14, "end": 407.14, "text": " So here's what I'd actually do with this if I were you. If you're not technical at all,", "tokens": [50365, 407, 510, 311, 437, 286, 1116, 767, 360, 365, 341, 498, 286, 645, 291, 13, 759, 291, 434, 406, 6191, 412, 439, 11, 50565], "temperature": 0, "avg_logprob": -0.05066190844904767, "compression_ratio": 1.681547619047619, "no_speech_prob": 2.28860420198318e-12}, {"id": 76, "seek": 40314, "start": 407.14, "end": 412.65999999999997, "text": " don't worry about downloading anything yet. Just know a model like this exists and what it can do,", "tokens": [50565, 500, 380, 3292, 466, 32529, 1340, 1939, 13, 1449, 458, 257, 2316, 411, 341, 8198, 293, 437, 309, 393, 360, 11, 50841], "temperature": 0, "avg_logprob": -0.05066190844904767, "compression_ratio": 1.681547619047619, "no_speech_prob": 2.28860420198318e-12}, {"id": 77, "seek": 40314, "start": 412.65999999999997, "end": 416.82, "text": " because it will show up inside more tools you already use over the next few months.", "tokens": [50841, 570, 309, 486, 855, 493, 1854, 544, 3873, 291, 1217, 764, 670, 264, 958, 1326, 2493, 13, 51049], "temperature": 0, "avg_logprob": -0.05066190844904767, "compression_ratio": 1.681547619047619, "no_speech_prob": 2.28860420198318e-12}, {"id": 78, "seek": 40314, "start": 416.82, "end": 420.65999999999997, "text": " If you do run a business and want to try it, the simplest place to start is feeding it a real", "tokens": [51049, 759, 291, 360, 1190, 257, 1606, 293, 528, 281, 853, 309, 11, 264, 22811, 1081, 281, 722, 307, 12919, 309, 257, 957, 51241], "temperature": 0, "avg_logprob": -0.05066190844904767, "compression_ratio": 1.681547619047619, "no_speech_prob": 2.28860420198318e-12}, {"id": 79, "seek": 40314, "start": 420.65999999999997, "end": 426.09999999999997, "text": " problem you have this week, like sorting through customer messages or summarizing notes and see how it", "tokens": [51241, 1154, 291, 362, 341, 1243, 11, 411, 32411, 807, 5474, 7897, 420, 14611, 3319, 5570, 293, 536, 577, 309, 51513], "temperature": 0, "avg_logprob": -0.05066190844904767, "compression_ratio": 1.681547619047619, "no_speech_prob": 2.28860420198318e-12}, {"id": 80, "seek": 40314, "start": 426.09999999999997, "end": 430.97999999999996, "text": " handles it. And if you want someone to hand you the exact setup instead of figuring it out alone,", "tokens": [51513, 18722, 309, 13, 400, 498, 291, 528, 1580, 281, 1011, 291, 264, 1900, 8657, 2602, 295, 15213, 309, 484, 3312, 11, 51757], "temperature": 0, "avg_logprob": -0.05066190844904767, "compression_ratio": 1.681547619047619, "no_speech_prob": 2.28860420198318e-12}, {"id": 81, "seek": 43098, "start": 430.98, "end": 434.5, "text": " that's exactly what we build inside the iProfit boardroom every single week.", "tokens": [50365, 300, 311, 2293, 437, 321, 1322, 1854, 264, 5180, 340, 6845, 3150, 2861, 633, 2167, 1243, 13, 50541], "temperature": 0, "avg_logprob": -0.11155609161623063, "compression_ratio": 1.651006711409396, "no_speech_prob": 2.1842560312540504e-12}, {"id": 82, "seek": 43098, "start": 435.06, "end": 441.06, "text": " That's the real story behind Qethos 27b, not hype, a small team closing the gap on what used to only", "tokens": [50569, 663, 311, 264, 957, 1657, 2261, 1249, 3293, 329, 7634, 65, 11, 406, 24144, 11, 257, 1359, 1469, 10377, 264, 7417, 322, 437, 1143, 281, 787, 50869], "temperature": 0, "avg_logprob": -0.11155609161623063, "compression_ratio": 1.651006711409396, "no_speech_prob": 2.1842560312540504e-12}, {"id": 83, "seek": 43098, "start": 441.06, "end": 447.22, "text": " belong to giant companies and handing it to anyone willing to use it. Before you go, two things. First,", "tokens": [50869, 5784, 281, 7410, 3431, 293, 34774, 309, 281, 2878, 4950, 281, 764, 309, 13, 4546, 291, 352, 11, 732, 721, 13, 2386, 11, 51177], "temperature": 0, "avg_logprob": -0.11155609161623063, "compression_ratio": 1.651006711409396, "no_speech_prob": 2.1842560312540504e-12}, {"id": 84, "seek": 43098, "start": 447.22, "end": 452.74, "text": " if a model like Qethos 27b feels like something you want running in your business, but you're not sure", "tokens": [51177, 498, 257, 2316, 411, 1249, 3293, 329, 7634, 65, 3417, 411, 746, 291, 528, 2614, 294, 428, 1606, 11, 457, 291, 434, 406, 988, 51453], "temperature": 0, "avg_logprob": -0.11155609161623063, "compression_ratio": 1.651006711409396, "no_speech_prob": 2.1842560312540504e-12}, {"id": 85, "seek": 43098, "start": 452.74, "end": 459.22, "text": " where to even start, that's the whole reason the AI profit boardroom exists. We turn releases like this one", "tokens": [51453, 689, 281, 754, 722, 11, 300, 311, 264, 1379, 1778, 264, 7318, 7475, 3150, 2861, 8198, 13, 492, 1261, 16952, 411, 341, 472, 51777], "temperature": 0, "avg_logprob": -0.11155609161623063, "compression_ratio": 1.651006711409396, "no_speech_prob": 2.1842560312540504e-12}, {"id": 86, "seek": 46098, "start": 460.98, "end": 465.78000000000003, "text": " same week they drop. So you're not stuck guessing how to set it up. You get weekly coaching calls", "tokens": [50365, 912, 1243, 436, 3270, 13, 407, 291, 434, 406, 5541, 17939, 577, 281, 992, 309, 493, 13, 509, 483, 12460, 15818, 5498, 50605], "temperature": 0, "avg_logprob": -0.08406275794619605, "compression_ratio": 1.6549520766773163, "no_speech_prob": 3.079548335088722e-12}, {"id": 87, "seek": 46098, "start": 465.78000000000003, "end": 470.74, "text": " where you can ask about your exact business, a full library of ready to use prompts and a community of", "tokens": [50605, 689, 291, 393, 1029, 466, 428, 1900, 1606, 11, 257, 1577, 6405, 295, 1919, 281, 764, 41095, 293, 257, 1768, 295, 50853], "temperature": 0, "avg_logprob": -0.08406275794619605, "compression_ratio": 1.6549520766773163, "no_speech_prob": 3.079548335088722e-12}, {"id": 88, "seek": 46098, "start": 470.74, "end": 476.5, "text": " business owners already putting tools like Qethos 27b to work in their own day to day. Links in the", "tokens": [50853, 1606, 7710, 1217, 3372, 3873, 411, 1249, 3293, 329, 7634, 65, 281, 589, 294, 641, 1065, 786, 281, 786, 13, 37156, 294, 264, 51141], "temperature": 0, "avg_logprob": -0.08406275794619605, "compression_ratio": 1.6549520766773163, "no_speech_prob": 3.079548335088722e-12}, {"id": 89, "seek": 46098, "start": 476.5, "end": 482.02000000000004, "text": " comments and description if you want the full walkthrough. And second, if you just want the free version of", "tokens": [51141, 3053, 293, 3855, 498, 291, 528, 264, 1577, 1792, 11529, 13, 400, 1150, 11, 498, 291, 445, 528, 264, 1737, 3037, 295, 51417], "temperature": 0, "avg_logprob": -0.08406275794619605, "compression_ratio": 1.6549520766773163, "no_speech_prob": 3.079548335088722e-12}, {"id": 90, "seek": 46098, "start": 482.02000000000004, "end": 488.74, "text": " that, grab a spot in the AI success lab. It's completely free. It comes with the notes from this exact video,", "tokens": [51417, 300, 11, 4444, 257, 4008, 294, 264, 7318, 2245, 2715, 13, 467, 311, 2584, 1737, 13, 467, 1487, 365, 264, 5570, 490, 341, 1900, 960, 11, 51753], "temperature": 0, "avg_logprob": -0.08406275794619605, "compression_ratio": 1.6549520766773163, "no_speech_prob": 3.079548335088722e-12}, {"id": 91, "seek": 48874, "start": 488.74, "end": 494.34000000000003, "text": " plus over a hundred other AI use cases you can start using right away. There's a community of more than", "tokens": [50365, 1804, 670, 257, 3262, 661, 7318, 764, 3331, 291, 393, 722, 1228, 558, 1314, 13, 821, 311, 257, 1768, 295, 544, 813, 50645], "temperature": 0, "avg_logprob": -0.09076464176177979, "compression_ratio": 1.3161290322580645, "no_speech_prob": 2.6141738513479895e-12}, {"id": 92, "seek": 48874, "start": 494.34000000000003, "end": 500.26, "text": " 87,000 people in there already applying this stuff to their own businesses every day. Links for both", "tokens": [50645, 27990, 11, 1360, 561, 294, 456, 1217, 9275, 341, 1507, 281, 641, 1065, 6011, 633, 786, 13, 37156, 337, 1293, 50941], "temperature": 0, "avg_logprob": -0.09076464176177979, "compression_ratio": 1.3161290322580645, "no_speech_prob": 2.6141738513479895e-12}, {"id": 93, "seek": 50026, "start": 500.26, "end": 503.14, "text": " are in the comments and description. See you in the next one.", "tokens": [50365, 366, 294, 264, 3053, 293, 3855, 13, 3008, 291, 294, 264, 958, 472, 13, 50509], "temperature": 0, "avg_logprob": -0.270352139192469, "compression_ratio": 0.9838709677419355, "no_speech_prob": 3.6121633028435296e-12}], "language": "en"}