Transcription
Greetings. Today, we have a very, very curious analysis lined up. We watched a video on YouTube, and there the author, you see, claims: "You take a company's job openings, run them through [a tool], and bam, in 10 minutes you understand their entire strategy." It sounds, well, almost like a spy movie, doesn't it? Yes. Hello. This is truly intriguing because before, well, job openings were like background data, analyzing them was long and tedious, but now these large language models, LLMs, they are, like, changing everything. Job openings are becoming almost an open book. The main thing is to ask correctly. That's exactly what we're talking about. Let's try to figure out today how this even works, why it's become possible right now. And what hidden things can be extracted? The video author analyzed the company Anthropic. They are the ones doing it. So let's see what we can dig up about their job openings using neural networks. Well, shall we dive in? Let's first recall how it was before. You want to understand a company's strategy from its job openings? What do you do? Oh, before, it was a whole story. You had to manually, well, literally sift through hundreds, if not thousands, of these ads. I can imagine, sitting there, classifying, looking for common traits, weak spots, comparing them with products, a nightmare. Prompt engineering, as it's called. Artificial intelligence doesn't just collect data for you, it also analyzes it and draws strategic conclusions, something that used to take a ton of time. Moreover, it analyzes not just how many people they are hiring, but the very essence, what kind of roles, what are the requirements, how they are connected. And the video author shows exactly this, for example, with Anthropic. He used some special application called Lovable, I think. And the results, I must say, turned out interesting. What did they manage to find out about this company? Yes, the analysis through [the tool] gave several curious, so to speak, insights into Anthropic. Firstly, product strategy. It seems that the company is, well, doubling its efforts in developing its core model. This is evident. But at the same time, it's striking that there are, uh, few new job openings for platform engineers, well, those who build and maintain the core infrastructure, there are somehow not enough. Not enough. That is, well, relative to other roles. This might indicate that they are currently more focused on scaling what they already have, rather than creating fundamentally new platforms right now. And this, by the way, quite aligns with what Anthropic has been saying publicly lately. Uh-huh. So they are, like, expanding the application of existing things, rather than building something completely new from scratch. Logical. It seems so, yes, judging by the profiles they are looking for. The second important point that was highlighted is their serious attitude towards ethics and safety. The job openings directly feature, you know, specific roles. For example, specialists in Alignment Science. Alignment Science. That's something about aligning with human goals. Yes. Yes, exactly. The science of how to make [AI] safe and aligned with our values. And it even mentions model welfare. Well, it can be translated as the well-being of the model. So these are not just PR words, these are real positions they are looking for people for. This emphasizes that they are truly investing in solving these complex ethical issues. Indeed, this definitely sets them apart from many. What about commerce? How do they plan to sell their technologies? And here, the job opening analysis shows a very clear tilt towards the corporate sector. Well, B2B. Many positions are related to working with large clients, especially with startups. And they are looking for many B2B marketers. It's clear that they want to quickly establish themselves in the business solutions market. A real, you know, push. Understood. But often the most interesting thing is not what is there, but what is missing. The video author specifically emphasized this. What conclusions can be drawn from the fact that some job openings are not visible at Anthropic? And here, yes, the analysis becomes particularly telling. For example, the AI noticed that there are few job openings for so-called Sales Engineers. Sales Engineers. These are specialists who help sell complex technologies because they understand them deeply themselves. Tech salespeople, roughly speaking. Well, something like that, yes, and there are almost no mentions of post-sale technical support. What could this indicate? Hmm, likely, their B2B direction is still at, you know, a rather early stage of maturity. So they are actively looking for those who will sell. But the team that will help with the technical part of the deal and then support clients might not be fully formed yet or is already overloaded. This is a perfectly viable hypothesis, yes, based purely on hiring data. And for Anthropic's competitors or for providers of some related services, this could be a signal. Perhaps their sales team currently lacks resources for deep technical elaboration or support, and this is a potential window of opportunity. One can offer them partnership or their services. Interesting, and if you look at it from a job seeker's perspective? I'm looking for a job, I'm looking at Anthropic's job openings. What signals can I pick up? For job seekers, there's also something to think about. Firstly, as I said, there are few internships and entry-level positions. It seems they are primarily targeting experienced specialists. Secondly, this situation with platform engineers, whom they are not hiring many of, could be, you know, an indicator of potential risk. The risk of accumulating technical debt. Technical debt is when the infrastructure cannot keep up with the growth in load. Yes, exactly. When you need to run forward and there isn't always time or resources to maintain and develop the foundation. And the video author, by the way, connects this with recent reports of service outages. There were such reports. Wow. So the job opening analysis could have hinted at this vulnerability in advance, so to speak. Indirectly, yes. It points to a possible bottleneck specifically in infrastructure scaling. And from this, one can assume, won't Anthropic soon need even more investment and people precisely for developing its platform? Highly likely. And did the AI see any other potential, well, not weaknesses, but areas for development? Yes, the system also indicated possible gaps in positions, for example, product managers, those responsible for the product itself, its development. There are also few testers, QA, and customer support people. The video author suggests that this might be a legacy of their research roots. Well, when the main thing is the technology first, and then its packaging and service. And an important point, you mentioned, it doesn't just say things like that, it shows why it decided that way. Yes, and this is absolutely a key point for trusting such tools. The application used in the video provided links to specific job openings, explained the logic of the conclusions. It's not a black box. Of course, it's not perfect, it can make mistakes, but the analysis process itself becomes transparent, and the volume of information it provides in a structured form is enormous. Okay. With the example of Anthropic and this special tool, it's clear, but what should we ordinary mortals do? Who don't have access to such applications or don't want to bother with API keys, digital passes, and all that. Can something similar be done with more accessible tools like ChatGPT? Absolutely, and the video author shows this too. A similar analysis can certainly be performed using publicly available tools. Like Perplexity or ChatGPT. The main thing here is to formulate the query correctly. The prompt. He provides examples of such prompts for Anthropic. And how much do the results differ if you use, say, Perplexity instead of Lovable? Well, let's look at the example prompt for Perplexity, which was tailored for job seekers. The output format is slightly different. Perplexity first shows the sources, links to job openings, and then the conclusions themselves. The emphasis, naturally, shifts to career-related aspects. It mentions the rapid growth of their offices in Seattle and New York, very high salaries, well, that's not surprising for a top company. It also confirms the high competition for positions and that they focus less on consumer-facing features. The AI generally aligns with the first analysis. Understood? So the essence is roughly the same, but the presentation and emphasis are slightly different. And if you give a different prompt, for example, with a focus on competitive intelligence, what then? Yes, the author provided such an example prompt for Perplexity, but from the perspective of competitive analysis or a product manager. And this analysis yielded more details specifically on product strategy. For example, information surfaced about work on CLD codec – this is their tool to help programmers, and about the importance of their internal computing platform MCP (Model Compute), apparently. That this is their competitive advantage, a technological barrier. Oh, interesting details. And did this analysis notice anything new about B2B compared to the first one? Yes, and this is precisely what shows the power of different prompts. The second analysis captured what the first missed. Job openings were noticed that are clearly targeted at specific industries: healthcare, fintech. This indicates not just B2B in general, but a more focused approach to individual market segments. It turns out that if you use different tools or even just different prompts for the same AI, you can look at the situation from different angles. The video author calls this obtaining a comprehensive 3D vision. By combining the results of several analyses from different tools or with different queries, you can assemble a more complete picture, see nuances that one highlighted and another missed. It's important not just to take the first result you find, but to compare and synthesize. And what else curious did this second competitive analysis reveal? It also noted interesting details about the organizational structure at Anthropic. For example, some signals against a rigid hierarchy. And potential risks too. For example, the possible presence of managers without already formed teams. This might hint at future reorganizations. The risk associated with the practice of acqui-hiring was mentioned. Well, that's when a company is bought mainly for its team. Plus, an important technical point was noted: dependence on Google's TPUs (Tensor Processing Units). This is specialized hardware for AI. For those engaged in competitive intelligence, such details are simply gold. Indeed, a lot of information can be extracted. We've analyzed the Anthropic example in such detail, but what is the main, broader conclusion from all this? What does it all mean? The key lesson that the author emphasizes in the video is: "Artificial intelligence makes entire classes of data accessible for analysis that were previously, well, either simply ignored due to processing complexity, or considered some kind of informational noise. Job openings are just one example, but a very telling one. And what specific opportunities does this open up? Well, for different people, specialists. Well, look, firstly, for competitive intelligence, analyzing a competitor's public job openings is now a simple and very powerful tool. No insiders needed, everything is on the surface, just take it and analyze it. Secondly, for investors, it becomes easier to verify whether the management's words about strategy align with actual hiring priorities. They say they are developing new directions, but they aren't looking for people there. Hmm. A reason to think and ask questions. Thirdly, for salespeople, analyzing a potential client's job openings is a way to better understand their internal workings, their needs, pain points, what technologies they use, how they are organized. This allows for a much more accurate and valuable offer, not just a shot in the dark. So it turns out that data that companies might not have particularly hidden before, because who would analyze it on a large scale, now becomes such a transparent source of strategic insights. Absolutely. Before, such analysis was either very superficial or required colossal manual effort, but now LLMs automate and deepen it. The video author gave another example, from a different domain. Analysis of public photos. Well, various selfies. Modern AIs can determine the geolocation of a shot with very high accuracy just from details in the photo. Landscape, signs, architecture. Even if all metadata is erased. That is, data that seemed completely harmless suddenly acquires new meaning. And what's particularly curious is that we are entering a world where the value and risks of publicly available information need to be truly rethought. What was just background yesterday can today become the key to someone's strategy or a source of vulnerability. And most importantly, as the author emphasizes, this is not some super-complex technology for select geeks or corporations. Yes, and this is a key point. With the right queries, with the right prompts, which the author, by the way, promised to share. Such strategic analysis becomes accessible to many. Whether it's a job seeker choosing an employer, a product manager studying the market, an investor verifying a hypothesis, or a salesperson preparing for a meeting, everyone gets the opportunity to gain valuable insights relatively easily and quickly. Large language models, they, like, democratize access to the analysis of these new layers of information. Well, today we've seen a rather vivid example of how AI can turn ordinary, generally publicly available job listings into a powerful tool. A tool for understanding a company's strategy, culture, even some of its weak spots. This directly shows how technologies are changing the very approach to information analysis. What used to take analysts weeks can now be done, well, figuratively, in 10 minutes. with the right query to a neural network. Yes, this is truly impressive and, you know, prompts reflection. And here arises a, hmm, important question, I think: "What other publicly available data, which we are used to ignoring, considering trivial, well, informational noise? What of them might hide such deep strategic value in our era of AI? What information streams in our work, even in everyday life, that we haven't paid attention to, could turn out to be unexpectedly significant if viewed through the lens of modern neural network capabilities? Excellent question. There's something to think about after our conversation. On this, perhaps, we will conclude our analysis for today. Thank you for being with us. Until next time.