Don t choose the wrong career in 2026

I was digging into some data last week and I went down a rabbit hole where I found that there's something very broken with how we train and teach new software developers. Markets, and that includes job markets, right, are supposed to be efficient. And I understand that every 10 years or so we do find ourselves in some sort of stock market bubble that makes it feel like that's not true. But I do believe that most markets are mostly efficient and everyone out there is looking for a good deal. And despite the well-known rationalities of human psychology, most of us do seem to do a good job of looking out for number one. That said, when it comes to which career in tech to choose, I've come to believe that the market is actually far from efficient. Why? Well, it's because learners are choosing career paths on hard mode. See, I assumed, and you probably have too, that the number of people trying to learn an employable skill would be roughly proportional to a few different things. Number one, the likelihood of actually landing a job in that field. Number two, the average salary for that skill. A better chance of employment means lower risk and higher salary means there's more reward for going into that field. So in theory, the best career path for any individual is the one with the lowest competition for jobs adjusted for salary. In other words, the one with the best combination of three different things. One, the most job openings. Two, the fewest people trying to learn it. And three, the highest salary. So here's a formula that we can use. The job score is equal to the job openings divided by search volume times average salary. And when I'm talking about search volume, I'm talking about the number of people searching for keywords that indicate they want to get into that field. So for an efficient market of eager learners, we'd expect new learners to prefer career paths with the highest scores. And because they prefer those paths, it would actually lower those scores on time, somewhat leading to a break-even score across all paths over time, assuming some sort of efficient job market. But when you dig into the data, that doesn't seem to be the case at all. Now, anecdotally, back in 2020, it bothered me that the number of people wanting to learn front-end development seemed to be much higher than the number of people wanting to learn back-end development. And it struck me as odd because on average, back-end developers do earn a bit more, and there are more overall back-end developers. And it was partly due to this perceived gap in the market that I decided to build boot.dev and teach back-end development there. And frankly, things have gone really well. See, I said to myself, it's not that fewer people actually want to learn back-end development. It's that there's just not that many online courses that are focused on teaching it. So fast forward to now in 2025, and we've shipped most of the back-end courses in our boot.dev roadmap. So this weekend, I sat down to do some research on where we should go from here. See, I get a lot of requests for other career learning paths, so I decided to try to quantify the demand for the various roles we're considering teaching material for. Among those, I'm considering back-end developer, obviously, we already teach that, front-end developer, full-stack developer, data engineer, DevOps engineer, data analyst, and AI engineer. Now, to be fair, what exactly an AI engineer is, no one really knows, but let's set that aside for a second. Now, for each role, I wanted to answer each of these questions so that I could plug the results into my formula. One, how many people want to learn these skills? Two, how much does the average person in these roles earn? And three, how many job openings are for these roles? So I cobbled together a few data sources. One, the Stack Overflow Developer Survey for 2025. Two, counts of job listings from the Job Data API. Three, Google search volume from the Google Ads Keyword Planner. And four, the Bureau of Labor Statistics. I'm just kidding, that one was actually basically useless. It just puts the data into overly broad categories like computer programmer, and that's not what I'm trying to find out here. So let's dig into what I found. Now, before we go on, I do want to put in a little disclaimer. You might be thinking, Lane, really, what is an AI engineer? And to that I say, great question. I don't know. And I don't think anyone else really knows either. But it does seem like it might mean some combination of training custom LLMs from scratch, or fine tuning existing LLMs, or prompt slash context engineering, building on top of LLMs, RAG, Retrieval Augmented Generation Systems, or just like using chat GPT really, really, really good. Again, we'll just kind of set that aside for now and talk a bit more about it later. Now, if you have a Google Ads account, you can access their Keyword Planner tool, which gives you the average monthly search volume for any given search term. The data is actually quite reliable and granular. The only challenge comes with identifying the best keywords to use for each role. I ended up using the following search terms, averaging the searches in the United States over the last 12 months. So from November 2024 to November 2025. So for back-end developer, I use the term back-end course. It had 50 monthly searches, which is only 0.24% of the total in this list. For front-end developer, I used front-end course, 1300 searches, full-stack developer, full-stack course, data engineer, data engineer course, DevOps engineer, DevOps course, data analyst, data analyst course, and AI engineer. I actually ended up using machine learning course, which it'll make more sense why I did that later when we talk about the jobs data. Now, as someone who sells interactive back-end courses for a living, this is actually a pretty scary table. It also explains why it was so easy to get boot dev to rank number one for the term back-end course on Google, but I digress. Now, before you freak out about this data, let me just throw in a few caveats and justifications for these keywords that I chose. AI engineer course had 1600 searches, but I think engineer is a bit too specific of a word to capture the machine learning crowd, which is clearly a big part of the AI pie, as we'll see in the jobs data later. AI course had 12,000 searches, but I think it's just way too broad. You see a bunch of people asking how to use chat GPT or mid-journey, which is really not what we're looking for here. And then software engineer course and coding course had about 5,400 searches each. And even though some of those searchers will likely go on to back-end roles, I think it's just too broad to be useful in our breakdown. I wanted to find a way to include terms like Java, Python, and Golang course in the back-end category, but I think it just muddies the waters even more than it clarifies them. So I left out all the technology specific terms, especially considering that something like Python course is far from specific to back-end development. Now, while these search terms are imperfect, they're my best attempts for now, so we're just going to roll with them, but keeping the limitations of the data in mind. Okay, so the next question is how much does the average person actually earn in each role? And this is probably the easiest question to answer with a high degree of confidence, thanks to the Stack Overflow Developer Survey for 2025. They ask respondents about their salary and break it down by role so we can get a good idea of what each role actually pays. Here's the 2025 data for the United States. Back-end developers earning $175,000, front-end developers, 145, full-stack developers, 138, data engineers, 150K, DevOps engineers, 165, data analysts coming in the lowest at 100K, and then AI slash ML engineers, the highest at $189,500. Now, aside from the looming question of what the hell is an AI engineer, this all seems pretty reasonable. Data analyst roles typically require less technical training than the others, so that disparity makes sense, and ML engineers actually need a lot of math, so again, it makes sense that theirs would be a little higher. Now, one thing that might jump out at you is that full-stack developers make less than back-end and front-end developers, but you need to know what full-stack really means in the industry for this to make sense. A lot of people that are like super new to tech assume that full-stack means like super goaded senior engineer that can do everything, but in reality, it usually means developer who works at a company small enough that they kind of have to do everything, and in that light, the pay disparity, again, kind of checks out. It's this jack-of-all-trades role. Okay, so for the next question, how many job openings are there for each role? And I dug around quite a bit trying to find some decent data on this. Frankly, I really wish that LinkedIn or Indeed or one of these large job platforms would just publish a big job openings data set or make it easier to aggregate the counts of job postings, but turns out that just doesn't seem to be the case. So I ended up using the Jobs Data API. By the way, everything in this video I'm going to link down in the description below. It ingests hundreds of thousands of job postings from around the world, and on this one page, you can see the number of postings over the last 30 days for each job title. Now, it's not great for looking at trends over time because we just have the last 30 days of data, but it does seem to be a pretty good snapshot of what's going on in November 2025. Here's a small snapshot of the raw data. As you can see, it's just broken down by job title. Now, the job titles don't nicely fit into the buckets because some of them are super specific with extra words and keywords at the beginning and the end, so we're going to have some cleaning to do here. So I manually downloaded all this data and manually classified all the applicable roles, which ended up being about 100 out of the total 2073, into the seven roles that we actually care about. For example, I classified Business Analyst as Data Analyst and Cloud Engineer as DevOps Engineer, or Senior Python Developer as Backend Developer. Now, admittedly, my classifications are not perfect. They're just the best I can do, but I am pretty happy with it, and I uploaded both the raw and classified data to a GitHub repo. So if you'd like to scrutinize my decisions, you can go ahead and do that. I'll link that repo down in the description below. Now, once everything's been aggregated, these are the counts that we get. Backend Developers had 1,162 job postings, which is about 6.25% of the total. Frontend Developers, 1,500, 8%. Full Stack Developers, 8,043%. Data Engineers, 1,500, about 8.3%. DevOps Engineers, a little over 1,600 for 9%, and Data Analysts, 1,900, a little over 10%. Finally, AI Engineers, 2,600, a little over 14%. Now, as you can guess, making a classification isn't always so cut and dry. So I tried to stick to a few rules. One, I removed managerial roles completely. Two, I removed any finance-specific roles. This one did kill a lot of potential Data Analyst roles. I removed hardware and embedded roles entirely. I removed overly generic roles like engineering manager. I also removed testing QA roles. I removed customer support and sales roles, and I removed product or project management roles. If a programming language was in the title, then I did my best guess at the role type. If it was too generic, I left it out. Okay, so with all those rules here, a couple things that I suspect from the final data. I think that backend openings are underclassified, and full stack openings are overclassified. For example, I classified Lead Software Engineer and Software Engineer 2 as full stack due to the generic nature of the titles, but I suspect that there are actually more pure backend devs in those titles than pure frontend devs. I also suspect Data Analysis is underclassified. I ignore roles that use Data Analysis but require additional training like Financial Analyst or Marketing Analyst. So I think we might be undercounting there. And we still have no idea really what AI Engineer is, but at least in this data set, it's mostly synonymous with ML or Machine Learning Engineer. That was what almost all of the records actually ended up being, which really should involve some training or fine tuning of models, which is why I think that the machine learning course search keyword was a good choice for us. Okay, now one bonus question I want to ask is just how many people are already working in each role? And this is a slightly separate question, but I thought it would be good for purely observational purposes and to point out another separate source of data. So Stack Overflow's poll shows the percentage of respondents working in each role, but Stack Overflow's community is biased really hard toward web development, which leads to some crazy numbers. In their data, backend developers make up 11% of the total, frontend developers only 4.3%, full stack developers 27.9%, and then Data Engineers, DevOps Engineers, Data Analysts, and AI Engineers are all like sub three or sub two percent. Now, I think this does a good job to show the relationship between backend, frontend, and full stack developers, but I really doubt there are four times more frontend developers in the United States than data analysts. It just seems more likely that data analysts simply don't hang out on Stack Overflow very much. So I'm going to mostly ignore this data except as another source to show the relationship between the different types of web developers, which I do think is fairly accurate. So time to answer the question you all came here for. Which skills should I be learning to maximize my chances of a high-paying job? Now, the real commercially safe thing for me to say here is, well, whatever you find most interesting and believe you can find a job for. After all, you only need one job. But of course, we're here to crunch some data, so let's actually quantify the options. Here, I've thrown up on the screen everything we've talked about so far, but smooshed all into one table so that you can review it line item by line item. We've got the percentage of job postings, the percentage of search volume, and the average salary of each role. Now, if you remember our formula from earlier, and I'm also going to divide by 1,000 to make the numbers a bit more manageable, here it is. Score equals job openings divided by search volume times average salary divided by 1,000. So we can now calculate the score for each role and then reorder the rows from best to worst. And when we do that, we get some pretty wild results. Back-end developer gets the highest score of 4,577. Full-stack developer, 788. DevOps engineer, 355. Front-end developer, 189. Data engineer, 109. AI engineer, 87. And data analyst bringing up the rear with 26. Now, that crazy low number of people searching for back-end course and the high number of people searching for machine learning course and data analyst course really does a ton of the heavy lifting at the edges here. So this data only holds up if you trust those search numbers. And while I trust the numbers themselves, the thing that I'm more skeptical of is how we actually found those search terms. I still don't have a better way to do it, but here we are. Now, you might be thinking, hold on, Lane. Isn't it super convenient that this table makes it look like everyone should be learning back-end development and you sell back-end courses on boot.dev? To that, I would answer yes, and you should probably be aware of that bias. But I will tell you that I really didn't set out to prove that back-end is a great choice because when I started putting this data together, really my goal was to find out which career path we should focus on teaching next. And we're actually already heavily invested in DevOps and data analyst paths. So in that way, this is actually somewhat inconvenient data for me because data analyst actually looks like maybe not the best choice for a lot of people. There are a lot of jobs, so I don't want to be all doom and gloom. So there's no way where you're getting a job as a data analyst person. But there's clearly a lot of people that are interested in learning data analysis. So, I mean, we're going to teach it. We're going to do our best job to teach it well. Just know that there is a lot of competition in that field, right? All that said, there is another variable that's not captured here, and that's how hard it is to learn these skills. Data analysis roles often require a lot less study and training overall, which helps to explain why those courses would be so popular with new learners, especially those who are looking for a quick career change. So I really don't want any of you watching this video to make any multi-year career decisions just based on this article alone. Please do your own research. The economy is in a really weird and crazy place right now, everywhere, not just in tech. And the only thing I can say for certain is that no matter which way you go about bettering yourself, don't stop bettering yourself. AI has obviously been making a ton of waves, and my biggest fear is that it's convincing entire swaths of people that learning stuff just isn't worth their time anymore. And frankly, that's just as false as it's always been. So please, don't panic. Good luck.