Chao Antigravity! Claude Code ahora es GRATIS y SIN Límites (Ollama) 💣 — Transcript

Full transcript

1. 0:00A tree can grow taller than all the others in the forest, but if it leaves its roots firmly planted,
2. 0:05the wind will eventually find its moment. Google Antigravity had become one of the
3. 0:10most used programming agents in recent months, largely due to its agent functions,
4. 0:15but especially because of its generous limits. Initially, it was limited, then they added
5. 0:20more extensive limits, but recently they reduced those limits. However, Antropic already had
6. 0:26Cloud Code, its code agent tool that consistently held the top spot,
7. 0:31despite only being available with its premium plans. Now, though, it has just been unlocked
8. 0:36to use its full potential without any subscription. I've been using it extensively in
9. 0:41recent weeks, and it's literally capable of doing the work of an entire team.
10. 0:49To prove this, we'll be exploring how to unlock it and testing it with
11. 0:53four use cases, from creating any type of agent, to connecting it with tools
12. 0:57like Nanobanana to create interactive websites that act as an analyst simplifying
13. 1:02data, or even as a complete team to create functional applications. The best part is that
14. 1:08you can use all its features today, completely free and
15. 1:14unlimited. Sounds good? Let's see how. To unlock all of
16. 1:21Cloud Code's free features, we'll need to combine it with a tool called Olama.
17. 1:26Founded in 2023 by just two people
18. 1:32with only $100,000 in funding, Olama has become a leading
19. 1:37open-source platform for running AI models both locally and on
20. 1:41servers. Let's jump into Olama, a tool with
21. 1:46a very clear focus: promoting building with open models.
22. 1:52Besides having a model section and a host of LLMs (Learning Modeling Tools) that we'll discuss later, Olama also allows you
23. 1:58to install various tools like Cloud Code, as well as many others, offering
24. 2:04over 40,000 integrations. Since we're going to integrate
25. 2:09Cloud Code in an unlocked way, the first step is to install Olama. To do this, we have two
26. 2:13options. First, we can copy this command from here, and in a new terminal,
27. 2:18if we paste and send it, the download and installation will begin. In fact,
28. 2:23it should only take about 10 seconds. Another way is to click this
29. 2:28download button, as they also have an application for Mac, Linux, and Windows. Once installed,
30. 2:33regardless of the method we choose, we need to copy this command, which
31. 2:38I'll also include in the description below. After typing it, we press
32. 2:42Enter, and that will configure the Cloud Code installation. Now that
33. 2:48it's installed, we just need to open it. To do this, we need to type this other command,
34. 2:52which essentially tells Olama to open Cloud Code. Pressing Enter
35. 2:57will then allow us to choose an artificial intelligence model. Here, as you can see,
36. 3:02we have some like Minimx, GLM5, the Quen models, as well as other recommended examples,
37. 3:08from which we can choose both cloud-based models that we use via API and even
38. 3:13models that we run locally. In fact, I already have some other models installed here
39. 3:17that we'll be looking at in the last use case, so let's start by using a
40. 3:22cloud-based model like GLM5. To do this, I press Enter, and from here we just have to press Enter
41. 3:28again to select "Yes" if we want it to have access to our entire computer, although
42. 3:33we can also give it access to specific folders. To do this, I could tell it not to access everything.
43. 3:38And after cleaning this up, before launching Cloud Code again, I'm going to type `cd` followed by a space and the path
44. 3:44to the folder I want it to have access to. In this case, I just opened the Downloads folder.
45. 3:48From the path, I copy the path, and if I paste it into the terminal, we'll be inside
46. 3:53the Downloads folder. And if I now tell him again to open Cloud from here by pressing Enter,
47. 3:59It will then ask you to select a model again, and once you do this, you'll have
48. 4:03access only to that download folder so you can make all the changes we request.
49. 4:10Starting
50. 4:13with the first use case, we'll be creating our own
51. 4:17artificial intelligence agents so that they can perform any task we need in the future. It's very similar to
52. 4:22hiring a person, except here we're creating the ideal
53. 4:27profile we need for a specific task. To do this, we could return to our
54. 4:31terminal and configure it from there, but instead of doing it from the terminal, I'll
55. 4:36also show you how to open it directly from an ID, such as Google Antigravity.
56. 4:41This platform, while it has its limitations, allows
57. 4:45us to install OLAMA with Cloud Code to use these advanced features
58. 4:51for free. From here, we gain access to the entire Wellravity interface.
59. 4:56From here, I could, for example, click on "open folder," select the download folder,
60. 5:02but now, within a project, instead of using the agent or Antigravity (
61. 5:06which we could close here), we'll click this other button to open the terminal. I'm already
62. 5:11seeing the right side because that's how I've configured it. Keep in mind that you could also do this
63. 5:15by right-clicking, selecting the position panel, and choosing the right side. Once you have this,
64. 5:20since we've already installed either Lama or Cloud Code, or you could also perform the same
65. 5:24steps directly from here, we'll essentially have the same terminal.
66. 5:28So, if I enter the same command as in Lama, like "launch clot," and press Enter,
67. 5:33the same options will appear, and I'll be able to use it in the same way, but with the added benefit
68. 5:38of previewing the files in the project. Now, continuing with this example,
69. 5:42let's create our agent. From here, if we click on the slash, we'll see
70. 5:47a bunch of options, including the second one, "Agents." If we press
71. 5:52Enter, it will appear that we can create our own agents. So, I'll
72. 5:56press Enter again. Now we can configure these agents to be used only in this project
73. 6:01or for more general, personal use. I'm going to create one specifically for this project.
74. 6:06In the next step, we could configure it manually, or, as I recommend,
75. 6:10let the cloud generate everything for us. So, let's press Enter on this first option.
76. 6:15Now we just need to describe what we'd like these agents to do.
77. 6:19As an example, I'll say, "I'd like an agent to help with the
78. 6:24data analysis I provide, which will start with an Excel or CSV file. The goal is for the
79. 6:29agent to analyze all that raw data and transform it into something more visual,
80. 6:35like a dashboard with charts, to help me make decisions quickly
81. 6:39." Okay, I'm going to leave that instruction as is. I've already pressed Enter to start
82. 6:44generating it. And notice how Cloud will start creating all the instructions so that
83. 6:50those people can have the best performance. A little later, it would ask us if we want to restrict access
84. 6:55to anything, like tools, editing capabilities, creation, reading, and so on.
85. 7:00In this case, I'm going to leave everything as is, so I'm going to click Continue and from here I'm going to tell it
86. 7:04to use the default model we're using. To do this, I click the fourth option,
87. 7:09tell it to automatically select a color, and notice that here we also have a
88. 7:12name, which would be "data dashboard analyzer". We could change this, but I'm going to leave it as is.
89. 7:17I enable the memory and after pressing Enter, we'll have the agent created. In fact,
90. 7:22if we go to the cloud section and select the agent section, we'll be able to see the new
91. 7:27agent that has been created for us here. If we click on it, we'll be able to see all
92. 7:31the detailed descriptions and instructions that Cloud has created to make this work
93. 7:36perfectly. Let's test it by uploading a document. In this case, I'm...
94. 7:41I've prepared an Excel file here that discusses the real estate market in Spain by province,
95. 7:46autonomous city, and autonomous community. In fact, in each of these tabs, we'll be able to see
96. 7:51all 50 Spanish provinces, the two autonomous cities, and the autonomous communities.
97. 7:57Each sheet contains very relevant data, such as territory,
98. 8:02property sales dates, price per square meter, per capita income,
99. 8:07the last rental date for which we have data, the average rent per square meter, and
100. 8:13many other data points from verified sources. Of course,
101. 8:18understanding all this data will take time, both to read and analyze. Now that
102. 8:22I've uploaded this Excel spreadsheet from my downloads folder,
103. 8:27by pressing Escape to go back, I can now click on the @ symbol and start
104. 8:33typing the name of my agent, which in this case would be the first one. So, I'll press
105. 8:37Enter to select it. And from here, if I give it some key information, for example,
106. 8:42to show me in a simplified and visual way the key points of the Excel file I just highlighted (
107. 8:47which would be the file name), and that for this I requested an HTML panel with dark mode. And
108. 8:52here, since people already have all the necessary information, we don't have to write
109. 8:56a very long prompt, because if I click send, Jacklot will start analyzing it, and
110. 9:02a little later it will ask for permission to run different commands and code.
111. 9:07So I'll accept it here. And a little later, look what we have on the first try,
112. 9:12with an application called Habitat 2026, showing the analyzed provinces,
113. 9:18the maximum average price, the top profitability, and along with this, we have different
114. 9:22graphs. Here we can see the geographical distribution with prices from highest to lowest,
115. 9:26and all of this is interactive. We also have the profitability versus the annual variation for each
116. 9:32of these provinces. We can also see the ranking of the top 15 by price. And ultimately,
117. 9:38I can now see all this data in a very expensive way, but I'll also be able to
118. 9:42filter it. If I wanted to see it by autonomous community, with a click, it
119. 9:47would start to change and I could quickly visualize it. The same would happen with the
120. 9:52autonomous cities, and if I go back to the province section, I'll have sales data, I'll also have
121. 9:57rental data, and I'll even be able to change the order, for example, by
122. 10:03gross profitability. Here we can see how the graphs have changed, just as if I, for example,
123. 10:07selected the highest annual growth, it would also change. With this,
124. 10:12we're no longer just talking about converting a lot of raw data into something visual; we're actually going
125. 10:17to be able to create our own agents, like a specialist that we can then
126. 10:22reuse for repetitive tasks. What we've just seen really opens up a lot of
127. 10:26use cases, like the ones we'll be looking at next. But another aspect you
128. 10:31might find interesting is how Olama, along with Cloud Code, can remain
129. 10:36active at all times, never shutting down when you turn off your computer. To achieve
130. 10:40this, you can install it on a VPS. I did this through Hostinger, since
131. 10:45clicking on the AI ​​and IML section there reveals a multitude of
132. 10:50artificial intelligence tools, including Olama. To deploy it, simply
133. 10:55click on the Docker Manager within the catalog section
134. 10:59, then click "Deploy," confirm, and
135. 11:04the configuration will begin automatically. In about 10 seconds,
136. 11:08it will be running. And if you open it,
137. 11:13you'll see that it's running on a cloud server,
138. 11:18ensuring that all your automations, tasks, and applications will always be available.
139. 11:22They remain active, regardless of whether we turn off our computer. To activate Olama,
140. 11:27keep in mind that I'm going to leave you a special link, this one here, which will
141. 11:31allow you to have everything installed automatically with just one click. You simply had to click on
142. 11:35this expandable section and then on the discount coupon section.
143. 11:40Thanks to Hostinger sponsoring this video, they've given me a coupon, which is
144. 11:45Alejand's, to apply, and you'll find an even better offer. I
145. 11:50'll leave the special link below in the description, and now that we know this, let's move on to the second use case.
146. 11:59Nano Banana 2 is the best AI-powered image generator, and Cloud Code is the
147. 12:05best tool for programming with agents. And if we combine both, we'll be able to
148. 12:10easily create highly visual and interactive professional websites. To see this use case, let's first
149. 12:15jump to Flow, this platform here, which has recently been updated, but the most
150. 12:21interesting thing is that we can use Nano Banana 2 for free, albeit with limitations, and with
151. 12:26advanced configurations. To do this, notice that when we click on the images section,
152. 12:31it tells us that each image will cost zero credits, so
153. 12:36we can use all 150 credits from the free account without any problem. With that
154. 12:40clear, I'll move on to the McDonald's website, as I'm going to try to improve
155. 12:46this site, which is probably worth several thousand euros. Going to the
156. 12:51burger page, I saw that while it's not bad, it does have some room for improvement. We can see
157. 12:55that there are static products, but beyond a few offers, the
158. 13:01main images, like this burger here, don't have any animation. So now
159. 13:05we're going to redo it with Cloud Code and Nano Bananana. To do this, the first thing I'm going to
160. 13:10do is [ __ ] this image. So I click on copy image, from Flow, click on paste, and
161. 13:16it will start loading from here. Once loaded, I'll specify it from the prompt section.
162. 13:22In this case, I want it to generate an image just like the previous one, but with all the ingredients
163. 13:27separated, as if the three hamburgers were floating in the air and the ingredients could be seen
164. 13:32individually. Since this is free, I'll select the option to
165. 13:36generate four images. I click send. Seconds later, on the first try, it generated
166. 13:41these four images, which, to be honest, for being on the first try and with such a complex program,
167. 13:46it did quite well. And honestly, of all these examples, I'd say this image
168. 13:51here is practically perfect, since we can see the ingredients, which are almost exactly
169. 13:58the same as those in the hamburgers here, only now we can see everything separately. So now,
170. 14:03with these images, I could [__ ] one of the images, drag it to the
171. 14:07ingredients section, choose the original image, drag it to the ingredients section as well,
172. 14:13and now where it says nano banana, we could change the video section. Here we do have a
173. 14:18written cost, so I'm going to reduce it to two videos with Veo 3.1 Fast. And since we've already configured
174. 14:24it to start with the ingredients separately and then finish with that final image, I'm going to tell it
175. 14:28from the prompt section to arrange the hamburgers until all the
176. 14:32ingredients are together. Click send, and a second later we might see this example here
177. 14:45, or the one I liked best, which we'll be using in this other video.
178. 14:58Once you have the result you'd like to put on the website, simply click
179. 15:02the three dots, select download, and from here I recommend selecting
180. 15:06720p. It's true that you could increase the quality, although sometimes it gets a bit saturated,
181. 15:11so it's a matter of trying different settings to see which one you prefer. In my case, I'm going to download it in 720p, and
182. 15:16now that we have the device, we're ready to create the improved page. To do this,
183. 15:20from Google Antigravity, we're going to click open folder, and in this case, I've just created one.
184. 15:24A new database called "web." I'll open it, and once we're here, we're going to
185. 15:30share the video we'd like to have on the website.
186. 15:35I've already uploaded it here, and then we'll click this symbol here to add a file. We're going
187. 15:40to create a skill, which I'll also explain in the description area.
188. 15:44I'll write scrollweb.md, MD, although you can change this name.
189. 15:50The most important thing here is to paste all the text I'll share into a document.
190. 15:55I've included a lot of information so we can have this
191. 16:01functionality where, when we add a video, it integrates perfectly with the
192. 16:05webpage, making it interactive and giving us a
193. 16:10much better visual experience than the original. So, now that we have our video and this skill
194. 16:15, we need to open our terminal again. We open Clot from here.
195. 16:21For this example, I'll select GLM5. I place it on the right, and from here, if I add something,
196. 16:27like, I want it to create a webpage to represent the three McDonald's hamburgers
197. 16:31with premium ingredients, and I want it to know that I've attached a video called Burger.mp4 MP4, and that
198. 16:37I want it to be included in an interactive and well-integrated way.
199. 16:42But if I now simply tell it to follow the skill,
200. 16:47I type @ and select this skill we just created called web scroll. When I press Enter,
201. 16:53it will start processing it, and a little later, on the first try, look what
202. 16:59we have here, as it says McDonald's Signature, the new collection. Three hamburgers
203. 17:04designed for discerning palates, premium ingredients, unparalleled flavor. And with this, if I
204. 17:09scroll down, I can see how letters appear, how the experience is redefined.
205. 17:14I keep scrolling down, and we can see how the hamburgers are being assembled. Here appears what would be
206. 17:19the first hamburger, called the perfected classic, with a price of 8.90. Here's where it all comes
207. 17:25together. Another 9.50 burger, and as we scroll down, we see other effects,
208. 17:32in this case with the irresistible 8.50 crispy burger. And finally, as we scroll down, we see
209. 17:37this final image with a few more effects. A call to action to check the time, and that's it,
210. 17:42we've reached the end. The best part is that we don't have a video, because if I
211. 17:46scroll up, we can really see how we can interact and how everything is
212. 17:50perfectly orchestrated as we scroll. For example, if I keep scrolling up,
213. 17:55the animation separates, but as I scroll down, it all comes together so we can see
214. 18:02the video perfectly integrated with the website. And with this, we're not just talking about creating
215. 18:07simple web pages; it's about creating real animations and a truly original experience.
216. 18:13It's truly amazing how much we can do now, and if you want
217. 18:17to have all the concepts well-structured and learn everything in a practical way to gain
218. 18:22a comprehensive and strategic vision of Generative AI, you should know that we've just
219. 18:27launched a new intensive artificial intelligence program so you can learn everything with
220. 18:32me over 8 weeks. This is a personalized and
221. 18:38close-knit training program, with a maximum of 25 participants per class. You'll have me directly in each class, not live,
222. 18:44but via video call, so we can interact and you can ask any questions
223. 18:48you may have. Furthermore, all classes, materials, and resources will be
224. 18:53available to you 24/7 for life. And at the end, you'll not only receive our own certificate,
225. 18:59but you can also optionally validate it and add a dual degree with
226. 19:03a Spanish university for credits. Regarding this course, keep in mind that below in the description
227. 19:08you will be able to see a lot of information with a URL that I will leave in the course
228. 19:12and along with this you will also be able to see reviews from other students who have already taken this step.
229. 19:22Now we're going to see how we can give our agent a bunch of abilities with
230. 19:26just a click. To do that, I've just opened another project called Cleaner, since
231. 19:31I want to create a more complex application that can actually help me avoid a subscription
232. 19:36I was considering. Specifically, it was this application that cleans
233. 19:42your computer by analyzing all the largest files, prioritizing them, and freeing up space
234. 19:47. But since this is a bit more complex, let's
235. 19:51give our agent more abilities. To do this, I've just opened Cloud again.
236. 19:56I'm going to expand on this a bit more. From here, if we click on the slash section,
237. 20:01you'll see an option scrolling down, which is the plugin option, although
238. 20:06you could also search for it directly by typing that word. I'm going to press Enter, and
239. 20:11from here, if we swipe right, we can enter the marketplace and click
240. 20:16Enter on the first option to add from the marketplace. By doing this, we'll be
241. 20:21able to copy any repository so that everything can be integrated. With that in mind, I'm
242. 20:26also going to share this cloud repository with you, which contains a wide range of skills. Here
243. 20:32in the skills section, we'll see many different types of skills, whether it's
244. 20:37for frontend development, PDFs, PowerPoint, website testing, and basically
245. 20:43any skill a complete team needs. To use this, all we have to do is
246. 20:47copy this URL from here, paste it where we left off, press Enter, and it will
247. 20:53start extracting all that information. Once it's done, select the second
248. 20:57option, "Document Skills." Press Enter, install it in the project, and then
249. 21:03you'll see this folder with everything installed. So now, if I add a prompt,
250. 21:08which I just pasted here, asking it to create a local application like CCleaner,
251. 21:13but called Clean Navi, and telling it to connect to the folder I select and analyze
252. 21:18all its contents, and specifying the functions I need—that it can analyze the files
253. 21:22to detect what junk I can delete and that I have a simple interface to manage it—when
254. 21:28I click send, after processing and creating it, it shows me a very clean screenshot asking me to specify the
255. 21:33path to the folder I want to analyze, so I'm going to use my downloads folder. I click
256. 21:38on analyze folder, and notice how it practically instantly measures all the files I
257. 21:44have, how much space they're taking up, and down here I can see, for example, the last video I uploaded
258. 21:50to my channel, which is over 8 GB. Ultimately, I'd have everything sorted from largest
259. 21:55to smallest. It would even show me duplicate files here; I
260. 21:59could even select these files from here, and it would show me all the space I'll be able to free up as
261. 22:06I select new files. In fact, notice how it would upload those gigabytes we had
262. 22:11selected. It could also have a filter for large files, another for temporary files,
263. 22:17and all the duplicates we have. And it could also show me old files older than 6 months,
264. 22:23if I have any. And to see if this works in other folders, I'm going to change the
265. 22:26path to, for example, this folder here, which, as you can see, has a lot of files, since
266. 22:31I was creating my own custom-made artificial intelligence agent, which, by the way,
267. 22:36if you'd like to see it, I already have in this video here. So now I'm going to copy this
268. 22:41path from here, paste it in this section, click on analyze, and notice how it instantly tells me all
269. 22:47the files it has scanned, in this case 142. It would show me those files, which are the ones I actually
270. 22:52have in that folder. Everything is sorted from largest to smallest. And again, if I were to
271. 22:58delete this to return to the downloads folder, I can see that I'm back here. Let's try
272. 23:03deleting a file, for example, the video. I click on "move to trash," confirm, and
273. 23:08from here it tells me that it's no longer on the system. Let's look at the downloads folder. And indeed it is.
274. 23:13That video wouldn't be here anymore. And now we're going to try deleting another file, like
275. 23:18the last thumbnail of the video I uploaded to my channel. I see it's called confi.jpg,
276. 23:24and it's right here, so I click, move it to the trash, confirm, and
277. 23:29that file is gone. With this, we can not only create
278. 23:34applications to try to monetize or create a new project, but we can literally
279. 23:38create an app to hire for our personal use. Only now we can customize it
280. 23:45with a whole team that has a lot of skills to develop it.
281. 23:54Finally, with the last use case, we'll see how we can use
282. 23:58locally installed artificial intelligence models to avoid any limitations, since everything
283. 24:03we've done so far has been [ __ ] cloud-based models that have certain daily limits.
284. 24:07In fact, if we go back to Laama, you'll see that in the upper right corner you'll find an account
285. 24:12that they ask you to create when you're going to use this tool. And going into the
286. 24:16settings section, you'll notice that, although it's free, it tells us that the
287. 24:22cloud-based models we're using will have a limit every two hours. In this case, I've used approximately
288. 24:264%, and we'll also have a weekly limit that will regenerate automatically.
289. 24:32This is great if we have a less powerful computer and want to run the
290. 24:37more demanding models. However, nowadays we also have smaller models that we can
291. 24:42install locally to avoid any limitations. To do this, look directly
292. 24:46at Olama; if we go to the models section, we'll find a wide variety of models.
293. 24:51It's true that, as their documentation states, when we use cloud code,
294. 24:55if we scroll down a bit, we'll see the recommended models, such as the Kimika
295. 25:002.5, the GLM 5, the mini max 2.5, and the Quen 3.5. These last four models are
296. 25:08cloud-based, and they also recommend other models that aren't, like
297. 25:13the Quen 3.5, which is one of the latest open-source models
298. 25:18and one of the most powerful. So, if we go back to laama, you'll see that we can
299. 25:22filter by the newest models. Here we'll find models like the Quen 3.5 and the GLM5.
300. 25:30Also, keep in mind that we might find open-source models from other companies, like
301. 25:34Google's Gemine, but for now, I'm going to select one of the most powerful, the Quen 3.5.
302. 25:41And notice that from here they tell us it's suitable for all these applications, and in fact, to run
303. 25:46and open this tool directly from a local model, we can simply type this
304. 25:50command. But first, we have to install it, and for that, we need to know which one you should install.
305. 25:54Keep in mind that you'll find different versions here with different
306. 25:59parameter sizes, and each one will require a minimum amount of RAM. Here, for example,
307. 26:04we have the 0.8 version, which requires 1 GB of RAM, and other much more powerful models that would require
308. 26:10more than 80 GB. So, ideally, you should choose one that's suitable for
309. 26:16your computer, and if you want to download one, you just have to select it.
310. 26:20In this case, I'm going to select this one, since I have others installed that I'll be showing shortly.
311. 26:24So, I'm going to copy the name directly from here. I'm going back to the terminal,
312. 26:29although I could actually do this from Antigravity, like the previous use cases.
313. 26:33And from here I'm going to type the word "olama pull" and paste the identifier
314. 26:39we just copied from the Lama page. We press Enter and it will start connecting.
315. 26:44And look how it will start downloading. In this case, it will take about 40 seconds. We're
316. 26:50just missing the last 5 seconds, but before anything else, keep in mind that you might not know which
317. 26:55model you can use for your computer. In that case, it would be as easy as taking a screenshot and asking
318. 27:00the GPT chat which one you can install by entering your device name. And from here,
319. 27:04Depending on the number of parameters, each one will have a different size. For example,
320. 27:09the smallest only takes up 1 GB, but we have other models that exceed 80 GB.
321. 27:14By the time I've finished telling you this, it would already tell me that I've successfully installed this model.
322. 27:19Now, if we open Clotama from here, you'll see that in the section below,
323. 27:24an additional model will appear, one with 2 billion parameters, in addition to the ones I saw initially.
324. 27:29To select one, I'll choose the smallest one
325. 27:33for this test. Pressing Enter will start us using it. Instead
326. 27:38of using a cloud service, which has certain limits—admittedly large, but still significant—
327. 27:43this method eliminates those limits because we're using local data. So,
328. 27:48from here, I can now interact with it just like the other models. For example, if I say,
329. 27:53"Hello, how are you?" It will start thinking about it right here, and in just a couple of seconds,
330. 27:58we'll have the answer. And with this, wherever we are, for example,
331. 28:02in Antigravity, if we launch it, all these models will appear again. And if
332. 28:07I open the most powerful one I have, which in this case has 9 billion parameters,
333. 28:11it will already be loaded here. And if I write this prompt and send it, after
334. 28:16almost a minute, I wouldn't have written it here anymore, speaking in this case of the lobster,
335. 28:21but keep in mind that it took me a little longer because I have quite a few
336. 28:25tabs open, screen recordings, and so on. The truth is that this unlocks
337. 28:30many opportunities. Opportunities that we no longer have to wait for because now we can create them.