229: i asked gpt-6 astra how to become a data analyst (it was wrong)

If you haven't heard, OpenAI just released their latest model, GPT-6 Astra, and it's supposedly really insane. This isn't just some small update. Sam Altman, the company's CEO, says that we're now in the AGI era. It's supposedly amazing, and everyone's freaking out. But how good is it actually? I wanted to test out for myself firsthand, so I logged into my ChatGPT account and asked a simple question: how to become a data analyst in 50 days? So let's see how it does. But before we do, just a quick note. If you're trying to become a data analyst and want some personalized advice, I'll actually make you a custom data analyst roadmap for free. All you need to do is go to dataanalystroadmap.com, fill in a few questions, and it will literally give you a real roadmap for you to follow that's based on real intelligence, not just artificial intelligence. But let's go ahead and go to the artificial intelligent roadmap now. All right. Here's what it said. It said, "I'll put together a practical 49-day plan and check current learning resources to keep it focused on useful skills." Now, I'm actually doing this inside of what's called Codex, which is basically ChatGPT's coding version. And the reason I'm doing it inside of Codex is I only had access to GPT-6 Astra in, uh, Codex or in Work. And the problem with using Work is it has all the details about me and what I say in ChatGPT, so I, I first did it inside of Work, and the answer was really biased because it already knew a lot about me and what I suggest in the SPN method, and I wanted to give it an unbiased opinion that doesn't have any information from me, at least from me actually using it. So here's what it said. "You can build a solid beginner foundation and a portfolio in 49 days." I agree with that. "Getting hired within that time is possible but depends on your starting skills, experience, and job market." I agree with that statement as well. "Assuming you're starting from scratch, plan for three to four hours, uh, daily, roughly 100 to 150 to 200 hours." I think that's about the right amount for my boot camp data analytics accelerator. I say you need to spend about 15 to 20 hours a week, so that's, what, like, three hours a day, essentially. Uh, and our program, uh, is eventually, like, this long, so that makes sense. "I'd prioritize Excel, SQL, and Power BI with basic statistics and communication throughout." I think that's a good plan so far. "Days one through seven, Excel and data basics, sorting, filtering, missing values, duplicates, dates, SUMIFs, COUNTIFs, XLOOKUP, pivot tables, and charts." I'm a little bit worried that this knows about my curriculum inside of the accelerator program because this is exactly what we cover in week one. "Finish with a clean data set and one-page report answering five business questions." That's like literally what we do, and then we publish it on LinkedIn. So this is-- I'm still worried that it knows who I am, but I asked and it said it didn't, so, uh, maybe this is somewhere deep in its memories. Two, SQL fundamentals. Okay, maybe this is a little bit different because I actually recommend going to Power BI or, in my case, Tableau next. Um, it's going to SQL. The order doesn't really matter. It's just I think Power BI and Tableau are easier. It's like a little bit l-less of a s- learning curve than SQL for most people. And I re-- I wanna like stack the wins, right? So that's why I personally, after Excel, go into one of, you know, Power BI or Tableau. It says, uh, days eight through 14, SQL fundamentals, select, where, order by, group by, having, case, aggregates, and handling null. That's all very solid. I 100% agree with that. Finish with 25 to 30 practice queries you can explain. Uh, I love this. Here's a fun fact for you. Did you know that the Google Data Analytics certificate, which probably a lot of you guys listening or, you know, watching have taken, that you only do 20 queries, 20 SQL queries in the entire certificate program. It-- the program is supposed to take six months, and you do q- 20 queries, and here you are having, you know, uh, ChatGPT-6 telling you, "You can do 25 queries in one week." And I 100% agree with that, and that's what we do inside of the accelerator program. So that's makes sense. Days 15 to 21, SQL analysis, joins, subqueries, CTEs, date calculations, and introductory window functions. Finish with an analysis of sales or customers using multiple tables. Uh, this is really good advice because I think you need to build the SQL fundamentals first, and then you kind of get into more complicated things. I think joins are pretty complicated. Um, so moving that to week two makes sense. CTEs and subqueries, that makes sense in week two. And then introductory window functions, I think that's exactly where I draw the line on how much SQL you should know to land your first data job. You should kind of know what a window function is, but you don't have to be perfect at them, 'cause a lot of jobs don't even use SQL, and a lot of jobs don't even use window functions. But you should be familiar with what they are, and if you needed to do one, you could, you know, you could create one with, with AI and double-check it. Um, I like that we're analyzing sales or customer's data. I think that makes sense. Um, that's one of the projects we do inside the accelerator program. Days 22 through 28, Power BI. P- Power Query. That's interesting that it has Power Query as, like, the first thing. Relationships, basic DAX, measures, filters, and chart selection. Oh, this is interesting. I don't think this is putting a big enough emphasis on the data visualization and chart selection. It's, like, really more focusing on, like, the data modeling and the data prep and, like, the data infrastructure. And a lot of the time, you're not gonna be in charge of that as a data analyst, especially as a junior data analyst. These are good things to know, but I just don't think I'd emphasize it. I would focus more on data visualization, 'cause making charts is gonna be part of your job as a data analyst pretty much no matter where you go. And so far, we're a month in, and we, like, really haven't done much with charts. Um, it says you finish with an interactive dashboard whose totals match your da- your source data. Okay, yeah, great, like creating a project in a dashboard. Um, I wanna emphasize all this "finish with," it's great to finish with these things, but if they just stay on your computer, that's not enough. You have to post them to a portfolio. They have to be public. There's so many different ways that you can build a portfolio, um, so many different platforms you can do. You can do your own personal websites on Wix or Squarespace or Carrd. You could use LinkedIn, you could use Substack. You could even use YouTube if you wanted to make videos. Um, I've created my own portfolio hosting platform called My Datafolio that you can check out. It'll have a link in the description down below. That's what I think is best, but, like, there's so many different options, you guys. My point here is your desktop, your downloads folder does not count as a portfolio. Make sure all of these things actually leave your computer and get out. Okay, uh, 29 through 35. First portfolio project. See, this is one thing I don't like as well with, with these suggestions. It's like, why is this not a portfolio project? Why is this not a portfolio project? Why is this not a portfolio project? You did all the work. Like, why not just turn it into a portfolio piece? Like, why do we have to do a separate portfolio piece? That's neither here nor there, I guess. Uh, days 29 through 35. Take a business question through cleaning, sequel analysis, visualization, and recommendations. A complete case study with queries, dashboard screenshots, and findings. Um, okay, this is kind of what I would call, like, more of, like, a capstoney project, where it's like you're combining maybe SQL and Power BI or Excel and Power BI or something like that. Uh, days 36 through 42, second portfolio project. Use a different dataset, work independently, and explain assumptions. You'll leave with a f- second case study and five-minute presentation. Um, okay. What type of presentation? Days 43 through 49, interviews and applications. What to practice: SQL exercises, spreadsheet tasks, explaining projects, resume tailoring. Finish with a portfolio link, a focused resume, and targeted applications. Okay, so, so far with this plan, it's, it's good. Uh, I think, you know, it's talked about different skills to learn. It's talked about creating projects and putting them on a portfolio, but it really doesn't talk about networking at all, which is a big part of landing a data job. You know, in order to land a data job, you have to follow the SPN method: learn the right skills, build projects and put them on a portfolio, and network like crazy. Um, and I think it put a really big emphasis on the skills. Um, secondarily with the project, but, like, really, what ends up getting most people hiring is the networking they're doing, and I will include updating your LinkedIn and your resume as networking. And I think we're not focusing on that enough, and I think that we're focusing on it too late. In my accelerator program, basically in week s- one and two, you'll work on your LinkedIn, and then on week three, you'll work on your resume. And that way, you can basically start applying for jobs once you're 21 days in because it takes a while to hear back, right? Um, you know, one thing I mentioned earlier is getting hired within that time is possible, but it depends on your starting skills, experience, and job market. I'll also say, well, if you're only applying for jobs on day 43, you're not gonna land a job by day 49. It takes about a week, even if you're going to be the winning candidate. It would take you a week, uh, at least to probably even get the first interview. "On day one, also review 10 relevant job postings in your target location. Record the current requirements so you can adjust this plan to the roles you actually want." Ah, this is kind of interesting. So this is saying, like, look at 10 roles and, you know, adjust this plan based off of those 10 roles. Um, I think that's decent advice. I think there's also some advice, sound advice in just, like, looking at 10 roles, seeing what they require the most, and trying to learn the things that are the easiest to learn and require the most. That's one of the recommendations I give is not to learn Python 'cause it's only required in 20% of data analyst jobs, and it's really hard to learn. It's a steep learning curve, right? So it doesn't make sense to spend a lot of time learning Python Use this daily routine. 45 minutes, learn one concept. 90 to 120 minutes, solve problems or build something. 30 minutes, check results and revisit mistakes. 15 minutes, explain one finding in plain English. I like that it's putting a good focus on actually explaining your findings and doing some of the reporting. One thing I mentioned earlier is we need to make sure we publish all of our projects, and when my students publish their projects, they have to actually write down what they did, why they did it, and what they learned from it, and, like, what recommendations they'd have to the business based off of their findings. And I think that's really important to do 'cause that's, like, the most important part of a data analyst job. If you just analyze data for fun and you don't actually say what's gonna... you know, what the business should change, then we're just wasting our time, to be honest. Okay. Along the way, learn percentages. Okay. Weighted average, mean versus median, outliers, sampling bias, and correlation versus causation. Practice check, practice checking row counts, duplicate keys, and totals, especially after joins. Uh, I think that's pretty sound advoice- advice. Keep your learning resources small. Excel, Microsoft's pivot table guide. Oh, man, this is where I think I'm gonna disagree here. Like, it's just pointing to the documentation from Microsoft Excel about pivot tables, which I think is just boring, to be honest. Um, I don't think it's, like, the best tutorial on pivot tables on planet Earth. Uh, also, does it give you any of the data that you're supposed to have? No. So you're just supposed to read this and not actually do it? That sounds so boring, and I don't think I would ever learn from doing that. So, uh, I would, I would ignore that advice here, and I would try to do some hands-on stuff. SQL Bolts provides interactive browser, uh, exercises for queries, joins, and aggregation. Okay, that's fine. Power BI, Microsoft's l- learns data analyst path cover prepping, modeling, and reporting data. Let's take a look at this Um, your path to, uh, being a data analyst. So you have get started with Microsoft, uh, Power BI, Power BI. Yeah, I'm biased obviously, but I don't love getting like a data analyst roadmap from like a provider like Microsoft, 'cause they're obviously going to be biased. It's like Google Sheets... Or sorry, Google Data Analyst certificate teaches you Google Sheets versus Excel, and I know they're very similar, but literally Excel is used, what? 90 times more, 70 times more, I think is the stat I have on that. So it's like you should learn, really learn Excel instead of Google Sheets. Um, and I don't really think like y- your training should be like, "Oh, you should learn Power BI 'cause we created Power BI," type of a thing. That's just my opinion. Um, that, that could be fine. For your first project, try Microsoft's financial sample workbook. I think that's gonna be boring. Investigate which products and markets drive sales and profits, how performance changes over time, and what deserves further investigation. E-portfolio project should show the business question and data source. I agree with that. Cleaning decisions and quality checks. I agree with that. Your queries or calculations. I agree with that. Three supported findings. I mean, three is an arbitrary number, but some supported findings. And then five, recommendations and limitations. And I do agree with that. Um, but I think you should have the recommendations, uh, and the findings up top as well, like in an intro section. You should BLUF, bottom line up front. Because if you're expecting a hiring manager or recruiter to scroll through or whatever, you know, go through, "Oh, this is the business questions and the source and the queries and all these things," before seeing the recommendations and the findings, that's just not gonna happen. So get the buy-in from them by having the findings and recommendations up front in your portfolio project. For this deadline, I postpone Python unless your target postings consistently require it. If they do, substitute some second time, second project time with Kaggle's Python and Pandas lesson. Um, I actually think Kaggle's lessons aren't bad at all, so good recommendation there, and I also recommend to postpone Python because it's only required in 20% of data jobs, and it's a steep learning curve. So that's a good advice. Your 49-Day Readiness Check. Can you take an unfamiliar dataset, clean it, query it, verify the numbers, create a useful chart, and explain a recommendation without following a tutorial? Use this as your benchmark for beginning applications. Oh, no. That was so good until the very end right here. "Use that as your benchmark for beginning applications." No, no, no, no, no, no, no, no, no, no, no. If, if you use that as a benchmark for applications, you're probably never going to apply for jobs because even me, who I've been doing this for 10 years now, of course I can take an unfamiliar data set, I can clean, I can query, I can verify the numbers, create a useful chart, explain a recommendation without following a tutorial. I can do that, right? But it's like, did I do it the best that I could have? I don't know. Like, it, it... I could always make it better, and I, I'm never 100% confidence in my actual analysis unless I've spent months on it. And like, this is just like such a bummer line for your confidence. Like, you're never going to feel ready to start applying for applications, so you really shouldn't have a benchmark for applications. You should just start applying for jobs and let the market tell you the benchmark, right? Because- The, a hiring manager or recruiter, you know, if they're trying to see if you can actually do this, they can't really tell from your resume or your LinkedIn or from your application. They would... They need to interview you or give you a case study to see if you can actually take a data set and analyze it on your own. They can't really tell if you can or can't from your resume or your LinkedIn. So until you get an interview, and you fail the interview, like the tech portion, you just failed, or they gave you the case study and you couldn't do it, I would not use this as a benchmark at all. Like, I think you should start applying for jobs ASAP, and if you get an interview, that's really good. If you fail the interview, okay, we move on and we get another interview, and we do better from the lessons we learned from failing the interview. Um, man, I think this would keep a lot of people stuck. Um, so I think that's really bad advice to, to really make that your benchmark. And in fact, a lot of junior data analyst roles, you know, especially the ones that maybe don't pay amazingly, like you're not expected to be a senior data analyst where you take an unfamiliar data set, you clean it, you query it, you verify all the numbers, you create a chart and explain a recommendation. You might just be a SQL monkey. You might just write SQL queries and put that under a report. Or you might just create useful charts. Like, you, maybe you don't need to clean the data set. Like, it's not... This is, this is something you need to be comfortable with eventually, but not, not before applying for jobs. Maybe not even before landing your first job. Okay? So I think this is really discouraging, and, uh, I would be kind of depressed. And if, if I were following this, I don't know if I would ever start applying for, for roles. Um, that would be, that would be kind of disappointing. So overall, I think the advice, it wasn't terrible, but it didn't like blow me away. I certainly didn't feel like, oh my gosh, there's AGI. It's here. It's teaching me everything. I thought it did a good job of focusing on the right skills, Excel, SQL, and BI, and saying, "Don't learn Python or R right now." But I think in terms of how to learn them, it was pretty darn bland and pretty dang boring. Like, who wants to read the Excel product manual? That's not learning, that's reading. For me, learning is hands-on. It should be doing. It should be building something. Which that actually brings me to my next point, which is the projects. I'm glad it mentioned projects and doing projects, but I don't get why you have to wait like weeks before doing your first project. Why not do it as part of the learning process, and like do it earlier if it's so important? I also didn't think it emphasized sharing these projects nearly enough. Like, share them on your resume, share them on your LinkedIn, share them on your portfolio. Just share them with your neighbor, like with anyone. It's just, like, a project isn't really useful if it's not shared, so I wish it would've emphasized that a little bit more. And lastly, I felt like it focused way too much on learning data analytics and not becoming a data analyst. And listen to that again. Learning data analytics and becoming a data analyst are not the same thing, or at least I don't think they're the same thing. One is about learning the technical, like, actual frameworks of analyzation and the tools to analyze, and the other is more of like a street smart, hack your way, the actual grind, the effort of getting a job. And I'd argue that that one is actually more important because in the end, the job is what gets you paid. And if you don't do the second one, you don't magically get paid. Like, if you're the best data analyst on planet Earth, but you don't apply for jobs, you don't have a resume, you're not gonna get paid. So it did an okay job, but not a great job. If you want an excellent, free, personalized data roadmap just for you based off of real-life intelligence and data and first principles, then go to dataanalystroadmap.com. You'll fill in a few questions, and then we will give you a personalized roadmap that you can literally follow to land your first data job. This one's made by real intelligence, not artificial intelligence. Remember that