Learning to Code in 2026: The Data
"Programming is cooked."
As we all know, for over a year, that's been the narrative. But I like to actually look at data, which tells quite a different (and more nuanced) story.
Just last year I wrote about how supply and demand are broken in programming education. But it's a new year, so I thought it was time for a fresh look at the data to see how "cooked" programmers truly are in 2026.
I pulled job postings, unemployment by age, unemployment by college major, and layoff numbers. Then I also pulled a bunch of our own Boot.dev data (1.5 million registered users btw) to see how our learners are actually behaving differently this year.
Now, you gotta know about my bias up front. I sell computer science courses! A strong developer job market (and even more importantly, strong sentiment around the developer job market) is good for my business, so I have every reason to tell you things are going well. That said, I'll do my best to be brutally honest when they're not.
Speaking of brutal, some of the job data is not great. But, in many cases, things are looking much better than the doomer-influencers would have you believe.
Let's get into it.
Software job postings are up 33% from the bottom
Indeed publishes a daily index of software development job postings, and sets the baseline at "100" (in relative terms) on February 1, 2020.
I think we all know this, but just to be clear: 2021 and 2022 were bat-shit insane. Software postings more than doubled (100 -> 233.8) in two years.
During that time, every senior developer I knew was looking at their new junior teammates who had spent eight weeks learning React and landed $80,000+ jobs, and wondering "Who the hell hired these people?!?".
I was against the mad rush to jam underqualified people into developer roles then, but I don't think I have to worry about that anymore, because I doubt we'll see that level of over-hiring again. C'est la vie...
All this to say, I don't think we should use 2021 as a baseline for "normal".
The bottom for software job postings hit on May 17, 2025 at 61.1, which is slightly lower than the worst week of the pandemic, and about a 39% decline from right before the pandemic.
But here's the good news. Postings have climbed 33% since then, back up to 81.6! October of 2026 averaged 79.4, which is about the same as August 2023. So we had a dip and return in postings since the end of the pandemic to now... which is not what I expected to see. Per the AI-saaspocalypse narrative, I expected 2026 to be a lot worse.
If you plot the rise of AI usage (or even just Anthropic's revenue) against these job postings, it's pretty easy to notice that they don't seem particularly correlated. If drastically increased AI usage was going to cause fewer job openings, I would have assumed we'd be at a lower number of open jobs in November 2026 than in May of 2025.
More postings doesn't mean more hires
To be fair, a job posting isn't a hire, and the hiring data is worse than the postings data. The best hiring data I could find for software engineers specifically stops at the end of 2025, but there's some other broader 2026 data we can use to kinda fill the gap.
LinkedIn's economists published a report on US software engineer hiring in February 2026. They track hiring through LinkedIn members who report a new employer in the month they start. Here's their hiring rate, where 1.0 is the 2016 average, up to Dec 2025:
All four lines peak in late 2021, when software engineer hiring was running at about double its 2016 rate. Then it slowed sharply from mid-2022 to late 2023 and leveled off.
Here's the same data, zoomed in on 2024 and 2025:
Look at the far right. The national, tech sector, and overall software engineer lines all tick back up to about 0.93 by December 2025... which is great! But the entry-level software engineer line does not. It ends at 0.78.
Interesting. We'll come back to that.
So what about 2026? LinkedIn hasn't updated the software engineer report, but they publish a quarterly AI Labor Market Update that uses the same hiring rate, compared to the same months a year earlier. It doesn't break out software engineers on their own. Instead, it groups them into "AI-augmented" occupations, which is LinkedIn's bucket for jobs where AI can do some of the work but not all of it, like "software engineer" and "data analyst":
| US hiring vs. a year earlier | Jan-Mar 2026 | Apr-Jun 2026 |
|---|---|---|
| AI-augmented roles, all levels | -1.9% | -6.5% |
| AI-augmented roles, entry-level | -8.9% | -9.6% |
| All US jobs, all levels | -6.9% | |
| All US jobs, entry-level | -7.6% |
So no, sadly hiring has not followed postings back up. Software postings from April through June 2026 were about 15% higher than a year earlier, but hiring in software-like roles was down 6.5% over the same stretch. And entry-level hiring in those roles fell faster than everything else, which is the same thing we saw in the December 2025 numbers above.
That said, LinkedIn's own take is that the overall slowdown is mostly economic uncertainty, not AI. It's hard to say for sure. I suppose we'll only know in hindsight.
Software postings rebounded, but not others
Okay, so software postings (not hirings) hit their low in May 2025 and have rebounded since, but the rest of the job market has not even seen the postings rebound. Just look at Indeed's all-jobs index: unlike software jobs, it's been basically flat (even down a bit) since May of 2025:
Again, that's not what you'd expect from the AI-doomsday narrative where everyone is preaching that programming is the job most negatively impacted by AI. Now, based on personal experience, I'm willing to believe that it's the most impacted as to how our day-to-day work has changed with AI, but this data does not suggest that software engineering is being erased.
Here's what I personally see as an employer. Out of my 15 employees, those with strong coding, problem-solving, and critical-thinking fundamentals get far more out of AI and agents than those who don't... after all, they're not trying to duct-tape together a Frankenstein mess of no-code-ai-enabled tools to get anything done.
But it's different for entry-level
So far we've seen the good and the bad - now here's the truly ugly. When it comes to entry-level postings, software development is dead last out of the 46 sectors Indeed lists. It has a smaller share of entry-level postings than even physicians and surgeons:
Unfortunately, they don't publish this by quarter (as far as I could find), but they do chart annual data back to 2019 for tech as a whole, which combines software development, data and analytics, and the two IT categories:
And while these categories are bucketed a bit differently, you'll notice that... it was always pretty bad for entry-level tech. Yes, it's gotten worse over the last six years: senior postings are up from ~48% to ~57%, while entry level is down from 8.3% to 7.1% and mid level from 43.6% to 36.0%. So, the trend isn't great, but we're talking about a roughly ~14% change for entry level and ~17% change for mid-level. And while mid-level is at its lowest point over the last 6 years, entry-level has actually rebounded, coming off a 6.5% low in 2022.
So if anything, per this data, it's mid-level devs that have had the biggest decline in postings, not entry-level. Fascinating.
Young people do not have the worst job market ever
The narrative on many subreddits is "zoomers aren't getting hired, we're cooked." But here's the data for all 20 to 24 year olds from 1990 to 2026:
About 7.5% of 20-24 year-olds were unemployed in October of 2026. From 2001 through 2016, there wasn't a single year where the average was that low. To be fair, 2025 averaged 8.3%, which was the worst year since 2021.
And, again, to be fair to zoomers, plenty of things are worse now than they were ten years ago. Housing costs take up much more of your paycheck. Gas is incredibly expensive. I could get a Beefy 5-Layer burrito for 79 cents in high school and they're ~$5 now.
There's also of course the issue of gig-work potentially screwing with the numbers. A lot of people DoorDash or drive Uber when they're in between jobs, and I'm just not sure how well that kind of stuff is controlled for in data sets like this, so it's worth mentioning.
But clearly there's a gap between these numbers and the "mood" online. My best guess? Algorithmic video feeds, and the "everything is terrible, you're screwed" angle gets more views than "young unemployment is only 7.5%" and that alone can honestly affect people's moods pretty significantly.
Recent grads lost their edge
This is the most interesting finding (at least to me) in this entire article.
For most of recent history, new college graduates have always had an "edge" over "all workers" when it comes to unemployment rates. However, per the recent grad data, that's no longer true:
For almost thirty years, a recent grad was less likely to be unemployed than the average worker. In 2009 the advantage was three full percentage points. When you look at a zoomed out version of this graph, it kinda looks like the "flip" happens during the pandemic. Which makes sense - a lot of things changed drastically during that period.
Fascinatingly, that's not true. The flip actually happened a couple of years earlier in 2018! They traded places a few times in 2019 and 2020, but then since 2021, recent grads have been worse off (in terms of unemployment) in every single year. And the gap is growing.
In September of 2026, recent grads were about a third more likely to be unemployed than the labor force as a whole. To be fair, they're still better off than other people their own age without a degree, as unemployment for all 22-27 year olds was 7.4%.
Two recent studies have looked into this strangeness:
In a June 2026 study, economists at the New York Fed estimated that remote work explains 64% of the increase in unemployment among young college graduates between 2017–19 and 2022–24. Their explanation: remote work makes it harder for managers to train and mentor new people. I imagine companies are reluctant to hire someone with no experience into a job where nobody is literally sitting next to them.
What about the other 36%? It could be employers losing trust in the education system to prepare graduates for careers in their field, but that's just my guess.
In a separate study, Stanford's Digital Economy Lab tracked ADP payroll records for 25,000 companies. Their August 2026 update says employment for 22 to 25 year olds in the most AI-exposed jobs is 19% below where it would be if it had kept pace with less-exposed jobs. Software developers are in the most-exposed group. Employment fell hard for developers aged 22 to 25, fell a little for the next age group up, and grew for everyone older. The authors are careful to call it a correlation, and they say the adjustment is happening through companies not hiring juniors, not through layoffs.
But in my opinion, neither study really explains why the lines first crossed in 2018. After all, 2018 wasn't particularly "more remote" or "more-ai" than the years previous.
Computer science majors have higher unemployment than art history majors
If you wanted a shocking headline, here it is. The New York Fed breaks the recent-grad numbers out into 73 college majors:
Computer science is tied for the fourth-highest unemployment of the 73 majors at 7.0%. Yes, 0.3 points above art history.
Computer engineering is second-highest at 7.8%, which is interesting to me because a year ago I kept hearing that computer engineering was the smart hedge against AI. "The AI boom needs hardware people to build GPUs and data centers..." yet it's doing worse than CS.
Okay, so unemployment isn't good for us software-learners... but not so fast. Let's look at earning potential and under-employment.
Underemployment is the share of grads working a job that doesn't typically require their college degree. For example, a barista with a bachelor's degree. They're not unemployed... but they're earning way less than they should be, and the regional Starbucks manager isn't reimbursing them for those student loans. Here's that data:
45.3% of art history majors are underemployed. For performing arts it's 63.9%.
Now look at computer science: it's only 19.1%, and for computer engineering it's even better at 15.8%. Only eight of the 73 majors have lower underemployment!
But not only that. Computer engineering has the highest early-career median wage of all 73 majors at $90,000, and computer science is second at $87,000. That's ahead of aerospace, chemical, and even electrical engineering:
And sorry to keep picking on art history, but it's sitting at $45,000.
So, yeah, a CS grad is about 2.8 points more likely to be unemployed than the typical graduate. But when you do a full risk-adjusted analysis for lifetime earnings based on what you study, it's undeniable that these fields are still way ahead of most, at least per this data.
I also have a personal anecdote that's stuck with me since late 2015. My dad dragged me to a church-related dinner, and a friend of a friend asked what I was studying. I said computer science. He said:
"Oh, that's what I studied. It's really tough out there. I got my degree six years ago and I'm still working as a custodian over at the hospital."
It really freaked me out at the time. I only had a ~year left before graduation. But as it turns out, I got a job lined up before I graduated, and never had trouble with employment since. I'm sure in many ways I was lucky, but my point is anecdotes (yes, even the ones with hundreds of upvotes on Reddit) are still just anecdotes.
These underemployment numbers aren't too unusual, if we look at history. Across all majors, recent-grad underemployment was 42.5% in September. In January 1990 it was 42.9%. It peaked at 48.4% in November 1992:
Remote work is shrinking
During the pandemic the pitch was that anyone could work from anywhere and get a job as a programmer. That wasn't true before the pandemic, and what do you know? We've been (very slowly) returning to that world. Indeed publishes the share of US postings that mention remote or hybrid work, by sector:
At its peak, 45% of software postings mentioned remote or hybrid work. It's now 31%, a third lower, and seems to still be falling. To be quite fair, that's 4x what it was in 2019, so it's not like "remote is dead". But this number includes hybrid, where you still have to live near the office.
Indeed also publishes a postings index for every US metro area. It covers all jobs, not only software, but you can see the change in total job postings from October 2025 to October 2026:
Change in job postings by metro area
All jobs, not only software. October 2025 to October 2026.
San Francisco's job postings grew faster over the last year than any other big metro I checked, and San Jose (Bay Area) wasn't far behind. Of course, they're climbing out of the crater left by the pandemic. San Francisco's index is still 28% below February 2020.
This is why I recommend new developers get an on-site job first if you can! You learn faster sitting next to real people, and when you apply for a local job you're only competing with the people in your city, not the whole world. Especially now compared to 2021, when 40% of software postings were remote or hybrid and climbing.
What changed among Boot.dev learners this year
Okay, so aside from just aggregating some Fed and Indeed data (which anyone can do), I wanted to share some unique numbers from our student metrics here on Boot.dev as well.
| Jan–Oct | 2025 | 2026 | Change |
|---|---|---|---|
| New account registrations (free) | 436,417 | 549,718 | +26.0% |
| New paid subscriptions | 51,690 | 55,753 | +7.9% |
| Solutions viewed | 2,214,996 | 2,114,359 | -4.5% |
| Chats with Boots (AI tutor) | 1,831,720 | 2,136,944 | +16.7% |
| Courses completed | 150,443 | 169,249 | +12.5% |
| Lesson resets | 2,066,197 | 2,890,132 | +39.9% |
549,718 people made a Boot.dev account in the first ten months of this year, up 26% from 2025. That said, we also spent 41% more on marketing, so it's not like our growth is directly indicative of the overall demand for "learning to code". But, it's also true that the audience hasn't gone away. And yes, advertising spend is up more than new customers, meaning we have to spend more per new student than we did in the past... which I have not enjoyed xD.
If we break down the intra-year a bit, you'll also see that paid subscriptions grew faster in early 2026, then fell below last year's pace towards the end of the year.
| New paid subscriptions | 2025 | 2026 | Change |
|---|---|---|---|
| January–April | 19,735 | 26,256 | +33.0% |
| May–October | 31,955 | 29,497 | -7.7% |
That said, from May through October, homepage visitors were up about 5%, signups were up about 4%, but new paid subscriptions were down 7.7%. We did run a particularly large sale in August and September of 2025 and didn't repeat it, so that affects the numbers, but we're also just noticing weaker conversion rates across the board this year. This is despite many A/B tests over the course of the last year objectively improving funnels and retention... which makes me lean toward weakening demand.
More new users start, but fewer keep going
This one is particularly interesting to me, as it speaks not just to "willingness to pay", but also overall motivation and focus in general. Here are cohorts of paid students, from both 2025 and 2026, and how far they get into our courses:
To make these cohorts apples-to-apples, each member is measured for exactly 90 days starting on the day of their first purchase, and we only counted Jan-July so we could get a full timing window.
What makes me sad to see is that, again, despite improvements to retention systems (we've shipped 20+ winning changes this year, measured with A/B tests) and the courses themselves (we've shipped literally thousands of updates to the course content, all based on learner feedback), we're actually seeing a dip in overall progress through the courses.
At the moment, I'm not sure what to make of this, but my worry is that larger trends could be affecting students, theoretically, things like:
Demotivation due to perception of worsening job prospects
Lower attention spans due to addictive short-form media consumption
Lower ability to "lock-in and grind" due to habitually leaning on AI to do more work
Solution views fell while tutor chats rose
On Boot.dev you can do two things when you're stuck:
View the instructor's solution
Chat with Boots, our (bear-wizard) AI tutor, who doesn't give out answers, but uses the Socratic method to nudge you along
Interestingly, chats with Boots have gone up against solution views pretty significantly, from a low of -26% to now +2.1% usage! This actually makes me very happy to see, and here's why.
While viewing a raw solution is probably a better learning method than having an AI "do your work" (because you still, at least some of the time, will read and understand the code) Boots chats are significantly better than viewing a solution directly, since he refuses to hand you explanations. He mostly asks follow-up questions, and draws your attention to things you missed.
Hopefully the data has helped shed some light on how the economy is always in flux, and while things aren't the best they've ever been, it's clear that they're also not the worst they've ever been. My hope is just that we as a species don't let the doom and gloom stop us from being curious creatures that love to learn and improve ourselves.
When I see a student give up on their coding journey, and pivot into medicine or the trades, it doesn't bother me. At least they're upskilling and striving to improve their lives. But PLEASE don't let macro-economic fears convince you that your time is better spent scrolling TikTok than it is getting better at something, whether that's software engineering or anything else.
Thanks, now try Boot.dev
Now, I wouldn't be running a very good company if I spent all this time on this article and didn't take an opportunity to plug our product! If you want to learn to code, or just want somewhere to keep learning so that your skills don't atrophy as you use AI more and more in your daily work, make a free account on Boot.dev! The first 3 chapters of every course are completely free, so you can try before you buy. Also, all the raw content is free, even beyond the first 3 chapters, so you can browse it to make sure it covers what you care about before you purchase.
Good luck out there.
Sources
Job postings index: Indeed Hiring Lab's job postings tracker, software development and all postings. Daily, seasonally adjusted, Feb 1 2020 = 100. Pulled Nov 10, 2026.
Seniority of postings: Indeed Hiring Lab, "The Labor Market Is Tilting Toward Seniority", July 23, 2026.
Job titles: JobData API titles page and Wayback Machine captures of it. Last year's classification is in this repo.
Search volume: Google Ads Keyword Planner, US, exact keyword, monthly averages.
Salaries and how developers learn: Stack Overflow Developer Survey 2025 and 2026 (30,903 responses, June 23 – August 5, 2026).
Unemployment by age and education: BLS Current Population Survey: ages 20–24, all workers, bachelor's and higher.
Recent grads, underemployment, and majors: Federal Reserve Bank of New York, The Labor Market for Recent College Graduates. Monthly series through September 2026.
Remote work study: Emanuel, Harrington, and Pallais, "Remote Work Leaves Younger Workers Sidelined", Federal Reserve Bank of New York, Liberty Street Economics, June 1, 2026.
AI exposure and young workers: Brynjolfsson, Chandar, and Chen, "Canaries in the Coal Mine?", Stanford Digital Economy Lab, August 2026 update.
Interest rates: New York Fed effective federal funds rate.
Hires: LinkedIn Economic Graph, "U.S. Software Engineer Talent Landscape", February 2026. The hiring-rate values in the charts above were extracted from the chart in the PDF.
Remote share and metro postings: Indeed Hiring Lab remote tracker and job postings tracker, data through September 30 and October 30, 2026.
