Someone asked me last week whether they should drop $18,000 on a coding bootcamp or just keep going with what they’re doing on YouTube and free Postgres tutorials. They’re 28, working in a non-tech job, and they want to be a backend developer in about a year. It’s the most common question in my DMs by a wide margin and the answer in 2026 is genuinely different from what it would have been three years ago. Let me try to lay it out without selling either side.

The bootcamp pitch in 2018 was simple: pay us, study full-time for 12 weeks, and you’ll come out with a job. That actually worked, more or less, for a couple of years. The market wanted bodies. Companies didn’t have hiring filters tuned for self-taught people yet. A bootcamp credential, even a so-so one, signalled “this person has at least gone through structured training.”

What’s changed is that the easy part of the job market — the abundance of “we’ll hire any front-end dev who can hold a conversation and use React” roles — is mostly gone. The remaining junior roles are scarcer and the hiring filters are tighter. A bootcamp graduate now competes against bootcamp graduates from the same cohort, plus self-taught devs with shipped projects, plus CS students, plus people pivoting from adjacent technical fields. That’s the part most bootcamp marketing pages still don’t tell you.

What bootcamps still actually do well

A good bootcamp gives you three things that are genuinely hard to give yourself. A schedule that you can’t ignore. Other humans in the same boat at the same skill level, which matters more than any tutorial. And someone to ask when you’re stuck on something for the third hour, which is the difference between learning and quitting.

If you’ve already tried to learn on your own and bounced off — if you’ve started three React tutorials and finished none of them — these things might be worth the money. Note that I said “might.” The variable is how good the specific bootcamp is, not bootcamps as a category.

What bootcamps don’t reliably give you any more is a job. The career-services teams are working harder than they used to, but the placement numbers most reputable schools publish are around the 60–75% mark within six months, and those numbers include people taking lower-paid roles than the marketing implied.

What self-study actually requires now

I learned mostly on my own. I want to be honest about what that took, because the version of self-study that works in 2026 is more demanding than the “watch some YouTube” caricature.

You need a structured curriculum, even if you wrote it yourself. Open up The Odin Project or freeCodeCamp’s curriculum and just commit to going through the whole thing. Skipping around is what kills self-study. The amount of material out there means it’s painfully easy to do six weeks of effort across nine half-finished tutorials and have no actual capability to show for it.

You need to ship things, not consume tutorials. The number of self-taught devs I know who’ve completed thirty tutorials and built two CRUD apps is much, much higher than it should be. Tutorials are the easy part. They are not the part that makes you a developer. Ship something tiny, then ship something slightly bigger, then ship something you’d actually want to use.

You need community. Self-study without other humans is brutal and, more practically, it leaves you with no signal on how you’re doing. Discord servers, local meetups, a few people you actually message regularly — any of those will work. None at all will probably not.

The honest comparison

Money: bootcamp $12k–$25k upfront or via income share, self-study $0 to maybe $300 in courses and tools.

Time: bootcamp 12–16 weeks intensive, self-study realistically 9–18 months part-time depending on how many hours you can put in per week.

Outcome quality: roughly comparable, with high variance on both sides. The best self-taught people are excellent. The worst are still confused after eighteen months. The best bootcamp grads are good. The worst spent $18k and are still confused after fourteen weeks. The single biggest factor in either path’s success isn’t the path. It’s how stubbornly you actually shipped real code instead of consuming content.

I’ve watched friends do both and end up in similar places. One did a bootcamp, hated it for the first six weeks, found a couple of people in the cohort who pushed her, and is now a solid mid-level dev at an Australian fintech. Another taught himself, spent his evenings building a side project that an actual small business started using, and got hired off the back of that. Both took roughly the same total time. Both got there.

There’s a third path I’ve started seeing more of: a hybrid where someone self-studies for six months to build the basics and confirm they actually like coding, then does a focused short program to fill specific gaps. I think this is the most rational version of the choice in 2026. You don’t pay tuition until you know you’ll finish, and you don’t pretend you can do it entirely alone.

If your question is whether the AI tools change any of this — they help on the margin, not on the fundamentals. Code review from a senior human still matters more than ten interactions with a model, and an AI consultancy I respect, Team400, recently published a take that mirrored my own experience: AI tools accelerate experienced developers more than they help beginners, because beginners can’t tell when the model is wrong. Use them, but don’t expect them to replace the work.

The real question is the one my DM correspondent should be asking themselves: which of the two paths can you actually finish? Whichever it is, that’s the right one. The other one is a way of spending money or time without finishing.