Transcription
Greetings to all. Let's start with a thought experiment. What if you could go to sleep as a solo freelancer and wake up as a team leader of fifty employees? >> It sounds like science fiction, >> absolutely. But it seems this is exactly what one new artificial intelligence tool promises us. Today, we're diving into Cwork. We'll try to understand what these creatures are and why there's so much talk about them. We'll rely on a very detailed breakdown from the YouTube channel AI Learning Communities. Its author, James, was one of the first to get access to this tool and, by all accounts, was immensely impressed. >> He even says there that this is the beginning of an era of agents for ordinary people. >> Yes, that phrase is what hooked me. >> And this definition for ordinary people, I think, is key here. Our task today is not just to list functions. We want to understand if this is truly a qualitative leap or, well, just another chatbot with a couple of new buttons. We'll break down this concept of digital employees, see how it works in practice with a live example, and, of course, try to peek into the future that this technology can unlock, especially for those who create, teach, or perhaps run their small businesses. >> Excellent. Let's go in order then. The first thing that immediately catches the eye in James's video is the interface. He specifically emphasizes that the button is not called "new chat," as we're all used to, >> but "new task." >> Yes, it seems like a small thing, but it's already a signal that the philosophy here is completely different. >> It's not just a change of name, it's essentially a change of the entire paradigm. We're moving from dialogue to delegation. That is, you're not chatting with AI, you're assigning it a task. There's a very precise thought in the video. Cwork is your first digital employee. >> Uh-huh. >> And most importantly, the quality of its work depends not so much on the AI itself as on you as a manager. >> That is, on how clearly you formulate the task, >> precisely how clearly you give instructions, provide resources, describe the result. All the responsibility shifts to us, as it were. So, it's no longer just a tool, but a team member who needs to be brought up to speed. Okay. And what can this new colleague of ours do? >> And this is where the most interesting part begins. >> James shows that Cwork can access files >> directly on your computer. >> And here, I think, many will immediately have an alarming thought about security. >> Well, yes, the first question. This is a key point, and the video's author emphasizes it. Access is granted manually to specific folders for a specific task and cannot just, you know, roam around your entire hard drive. >> Understood? So, we ourselves define its, let's say, workspace. >> Exactly. It only gets access to what it's explicitly allowed. >> But that's not all. It can also go online through a special Chrome extension, >> right? So, it's not confined to one tab. Its workspace is both our local files and the entire internet. It breaks down this digital barrier. >> Precisely. And the interface fully reflects this. There's not just a chat window, there are sections. Progress, you can see what it's doing. Artifacts – these are the files it has created. >> And context too, >> yes, context is the source materials you gave it, and connectors. We'll talk about those later. This is no longer a chat, it's essentially a project management dashboard. >> Okay, so it can work with files and Google. That's already good, but the real magic, as I understand it, begins further when we realize that this employee can be given ready-made instructions, almost like job descriptions in a real company. In the video, this is called skills and tools. >> Exactly. Essentially, a skill is a pre-recorded, very detailed prompt or even a sequence of actions for a specific task, >> for example, >> well, let's say, a skill for analyzing text sentiment or a skill for translating a legal document into a concise summary. You create it once, and then you just tell the agent: "Use this skill." >> Like giving an employee a manual? >> Exactly. Moreover, the author mentions that Google is already starting to support a similar file format for skills. >> Seriously? >> Yes. And this is already a hint that we might be on the verge of a new industry standard for training AI agents. >> To make all this sound less abstract, James shows an incredibly illustrative example in his video. He describes a problem that, I think, is familiar to everyone. >> Oh, yes. >> He calls it digital exhaust. Terabytes of information, notes, drafts that lie dormant. >> That very "to sort" folder that you never get around to. >> Exactly. For him, it's training materials, mind maps, scattered ideas for courses, a digital dump, in short. >> And so he decides to use Cwork to bring order to this chaos and create useful content. He sets the first very specific task. >> Generate topics for weekly discussions in my online community. >> Yes. Based on the materials from this folder. And he gives it access to the folder where files are in various formats. Markdown, mind maps. >> Text. The first thing it does is compile and show a to-do list. A to-do list. Yes, yes. Something like: First: analyze all files. Second: identify key themes. Third: group them. Fourth: write posts aimed at stimulating curiosity. >> It doesn't just execute the command, it shows its plan and, more importantly, its understanding of the goal. >> It understood not just "make posts," but "stimulate curiosity." >> That's a different level of comprehension. And then another magic happens. During the process, it also finds among its available skills a custom skill called "community feed publisher," >> even though it wasn't asked to do that. >> It wasn't asked to do that. The agent simply saw the task, analyzed the tools, and chose the most suitable one. >> Okay. And here, I think, the coolest thing in the whole video happens. While the first agent, this digital employee of ours, is sitting and working on the posts, >> analyzing files, thinking: >> "Uh-huh, James doesn't wait, he just opens a new tab and hires a second employee, >> gives him a completely different task. >> Transcribe a long YouTube video, this changes absolutely everything." >> Absolutely. It's no longer a conveyor belt where one task must finish for another to begin. It's already like a real office where several specialists work in parallel. >> One writes texts, another >> analyzes a report, a third prepares a transcription, an interview. That very idea of a team of fifty people in real time. >> I must admit, when I watched this part of the video, I had a real wow effect. >> I understand. It's one thing when AI performs one task, but when you see one employee continuing to work, and you immediately give a new assignment to another in a neighboring window, >> that's when you realize that this isn't just a chatbot, >> it's a multitasking workstation. >> Exactly. It's a transition from sequential task execution to parallel project management. Your computer turns into a hub where you, well, conduct and a whole team works. >> Okay, let's get back to the first task, let's finish it. The agent generated the posts, they appear in the artifacts section. James reviews them, he's satisfied. And then he gives a very simple command: "Publish the posts for the first day to the general feed." >> No "copy this text and paste it there." >> Just a direct instruction. And here's where those connectors we mentioned come into play. Essentially, they are digital keys that allow AI to go beyond its application and work authorizedly in other programs. >> So, it's like logging into the community platform. >> Exactly, on behalf of James, it finds the right section and publishes the generated content. >> And James shows the result on the screen. Three new posts appear in the community feed >> created and published by AI. Without a single click, without manual copying. >> Let's just stop for a second and absorb this. No copy-pasting. It understood the task, found the right tool, i.e., the skill, connected to an external website through a connector, and did the work. >> This is no longer an assistant who needs everything explained. >> It's an autonomous executor. >> And here we come to the main point. When you see such a demonstration, you start thinking about the consequences. This is not just a tool for geeks and programmers. >> Absolutely not. >> The entire philosophy of Cwork, judging by the video, is aimed at giving these superpowers precisely to ordinary people, >> entrepreneurs, creators, educators, people who have expertise but lack a team to scale it. Yes, a fundamental shift is happening. We stop being just content creators and become more like conductors of an entire content system. James's video describes his concept of a private branded campus. >> Uh-huh. >> The idea is that all your content, videos, articles, notes, courses, all of it is stored in a unified library, >> in one place. >> Yes, and the task of AI agents is not so much to generate something from scratch as to constantly refer to this library, find new connections, repackage knowledge into new formats. That is, instead of writing a new post every time, you tell the agent: "Go into my library, find everything about productivity, and make a series of five posts for LinkedIn from it." >> Exactly. And also a script for a short video. The goal is not to create individual materials, but to work with an entire ecosystem of accumulated knowledge. Your value is your unique database, and AI agents are the team that helps monetize and distribute it. Well, of course, a disclaimer is needed here. At the time of recording the video, as James says, the tool was in quite exclusive access. >> Yes, that's important context. >> Only for Mac, required app installation and a Max subscription, which cost $100 per month. >> Not cheap. And there was a long waiting list for the more affordable Pro plan, so it's not for everyone yet. >> It's a tool for enthusiasts, then >> for now. Yes. But what's interesting is that James offers advice even for those who don't have access to Cwork. He says you can start preparing for this new reality right now. >> How? >> In the regular Cloud app, under settings, there's already a "Skills" section. You can see examples there and even create your own. >> Ah, so start thinking like a manager and agents. >> Exactly. He even mentions that the developers at Anthropic have added a special skill called "Skill Creator." >> What is that? >> It's an AI that helps you create instructions for other AIs. >> Wow. >> So, you can start documenting your work processes right now, turn them into clear instructions, and save them as skills. Essentially, it's an opportunity to pre-train your future digital employees even before their official hiring. You write job descriptions for them. >> Exactly. >> And in the finale of his video, James expresses a very bold, almost provocative thought. He says that such tools open the way to a billion-dollar business managed by one person. >> Well, that certainly sounds like an exaggeration. >> Clearly. But the idea is understandable. >> The idea is understandable, yes. When you have a practically unlimited team of executors who can creatively process all your experience, production itself ceases to be the bottleneck. >> The main thing becomes your expertise. >> Your expertise and your ability to set tasks. The focus shifts from doing to conceiving and managing. If all the routine work of creating, adapting, and publishing content can be automated, then a person is freed up to focus on strategy and creativity. >> So, to summarize, we started our conversation with a simple question: is this a new chatbot or something more? And, in my opinion, the answer is >> more than. We've seen a transition from a conversational system to a task management system, from a single response to the parallel work of an entire team of digital employees, >> who can work with our local files and interact with the outside world. But, in my opinion, the main idea that the source conveys is not so much the excitement about the technology itself, but a call to change our own thinking. >> Uh-huh. We are invited to stop being executors and become managers and agents. The key skill becomes not the ability to write perfect prompts, but the ability to set clear goals, break down large projects into subtasks, and provide your digital team with the necessary resources. >> And in conclusion, as always, a thought to ponder that arises from our entire discussion today. Let's imagine for a second that this has become a reality. Uh-huh. >> A practically infinitely scalable team of digital employees has appeared. They have access to all accumulated knowledge, notes, files, drafts, >> all this digital exhaust. >> Yes. What is the one biggest, most ambitious task that has always been postponed due to lack of time or resources, what one large-scale project would you assign them first?