ChatGPT Hits 100M Users: What It Means for Juniors

Archive note: this is a backdated post, written years later while rebuilding this site. It’s dated to the moment it covers, but the hindsight is real.

Analysts are reporting that ChatGPT reached an estimated 100 million monthly users in two months — the fastest consumer app adoption ever recorded. TikTok took nine months to get there; Instagram took over two years. In December I predicted this thing would be in my daily toolchain within a year. Revised estimate: it took six weeks.

Two months of real use later, the question I keep getting from students and career-shifters deserves a straight answer: should I still learn to code?

What it’s actually good at (and not)

The honest inventory from daily use: it writes boilerplate, CRUD scaffolding, regexes, SQL queries, and test cases faster than any junior I’ve worked with. It explains unfamiliar code patiently and never sighs. It also confidently invents methods that don’t exist, produces Laravel code mixing API versions five years apart, and will defend a wrong answer with the serene assurance of a con man. The skill it cannot fake is knowing whether the output is right — and that, it turns out, is most of the job.

The junior question, answered honestly

Yes, learn to code — but the entry-level bargain is being renegotiated in real time:

1. The tasks juniors learned on are exactly what it automates. Fix-this-small-bug, write-this-form-validation, convert-this-spec-to-CRUD: that was the apprenticeship ladder. If seniors with AI do that work themselves now, where does the next generation practice? Nobody — me included — has a good answer yet, and the industry hasn’t noticed it owes one.
2. Verification becomes the core junior skill. The differentiating ability is shifting from producing code to interrogating it: reading critically, testing properly, knowing when the confident answer smells wrong. Ironically, that’s a skill we used to let juniors develop slowly. Now it’s the entry fee.
3. Fundamentals matter more, not less. The dev who understands HTTP, SQL, and data structures can catch the model’s beautiful nonsense. The dev who only ever prompt-and-pasted is a passenger in a car with no brakes. Abstraction without understanding has always been debt; this just lowers the interest-free introductory rate.
4. The floor rose; so did the leverage. One person who knows what they’re doing now ships like a small team. For PH devs competing globally, that cuts both ways — the same leverage is available to everyone, everywhere, tonight.

What I’m doing differently

Concretely, a month in: I draft tedious code through it and review like a hostile auditor. I use it as a rubber duck that talks back. I do not let it touch anything security-sensitive or architectural — the OWASP top ten doesn’t care how fluent the prose was. And I’ve started writing better commit messages and docs, because explaining context to the machine turns out to be the same skill as explaining it to colleagues.

The diff against every previous hype cycle I’ve covered on this blog: the tokens needed you to believe. This thing just needs you to have a tab open. Utility-per-week remains my metric, and the meter is running hot.

To the juniors: the door didn’t close. The doorway moved. Learn fundamentals, learn to verify, and learn the new tool — the people who fear it and the people who worship it will both be working for the people who check its output.