Advice by age
I am 26, so this is not wisdom. It is what I would tell myself at each stage, written down so I stop repeating it over coffee. The last section is empty on purpose.
under 20
Finish things other people can use
A tutorial teaches syntax; a finished project teaches design. Build a scraper, a bot, a small game, a tool for your school, and make it work end to end, including the boring parts: the config file, the error message, the README. Five finished small things are worth more than one ambitious abandoned one.
Learn the machine, not the framework
Memory, files, processes, HTTP, and one language's standard library. Frameworks you learn at 18 are worthless at 25; what a pointer or a socket is stays true at 40. If you can explain why a slow query is slow, you will never be lost in a new stack for long.
Get comfortable at a terminal
Not to look like a hacker, but because the terminal is where you will do your actual work: read logs, copy files between machines, kill a process, check what is listening on a port. Learn grep, ssh, less, and enough vim to survive on a server.
Read things that are not about code
History, philosophy, fiction. You will spend your career making decisions about people, and code is only the medium. Almost every interesting argument in software is an old argument from somewhere else wearing new vocabulary.
20 – 25
Learn one language to the bottom
Pick the language you will use at work, then read its standard library and, after that, its runtime. Knowing what net/http actually does, or how the Go scheduler treats a blocking syscall, changes the quality of your questions. Surface knowledge is cheap now; depth takes years and cannot be borrowed.
Read code before writing more of it
The fastest way to learn design is to trace one feature through a codebase you already use: one handler, one migration, one retry path. Ask why it is shaped that way. Most of what looks arbitrary turns out to be a scar from a real incident.
The unglamorous tools are the job
Terminal, git, SQL, HTTP, Linux. Frameworks come and go; these compound. If you can read an execution plan and explain why an index was ignored, you are already more useful than most people with more years than you.
Write in public, badly, early
One note a month is enough. You will discover what you do not understand the moment you try to explain it, and people will correct you for free. Over a few years those notes become a reputation that no line on a CV can buy.
Choose the people, not the logo
Your first two jobs should be chosen for the engineers you will sit next to and the problems you will be allowed to touch. A famous company with a bored team teaches less than an unknown one where someone reviews your code properly.
25 – 30
Judgment is the deliverable
Nobody pays for code any more; they pay for decisions about which code to write. Learn to state the trade-off out loud — what you get, what you give up, what would change your mind — before you start typing. That sentence is the difference between a mid-level and a senior engineer.
Own something end to end, including the 3 a.m. pages
The gap between good and senior is not knowledge. It is having been the person who has to fix it, explain it to the customer, and then make sure it does not happen again. Volunteer for that system. It is uncomfortable and it is the fastest education available.
Explain it until someone else can do it
Teaching compresses your own understanding more than reading does. Write the design doc, run the onboarding, review the code carefully enough to say why. If only you can operate the system, you are a bottleneck, not an expert.
Say no to work that does not compound
By 27 you have enough leverage to choose. Migrations, internal tools, and documentation compound; dashboards nobody opens and meetings that could be a paragraph do not. Every yes is a no to something with a longer half-life.
Protect the machine you run on
Sleep, a resting heart rate you know, and something physical you do badly on purpose. The career is forty years long; burning two of them for a promotion that disappears in the next reorg is a bad trade. I learned this one on a trail at kilometre 60, which is a late place to learn it.
30 and above
Not written yet. I am 26, and anything I put here now would be borrowed from someone else's life. I will write it when I have something true to say, not before.
If you disagree with any of this, email me. I answer faster than I should.