1. Identify a problem. (User wants do something and can’t, something that is supposed to work doesn’t, someone wrote shit code that works and we want to fix it)
2. Get more info about it: ask users for more context, find out about their workarounds, assess the impact of the bug, find solutions to similar problems. Get together with others and hash out some design.
3. Do the coding. Often involves a bunch of reading documentation and trial running code to see if it works
4. Come up with a way to confirm the change does what it’s supposed to: write a new automatic test, or a procedure a person can follow to verify it works
5. Write a description of the change and test plan
6. Get someone else to check what I’ve done and make any changes they ask for (as long as I agree)

I write code in a niche industry, in an even more niche language.

With 20 years experience I am finally at the point that much of my stuff works without too much headache.

Unfortunately, now that I’m finally good at it, it’s become a much smaller part of my overall job.

Nothing I look forward to more than being left alone for a few hours with my headphones on banging out a project.

10 Get told to stop working on whatever I’m working on, some tech debt just became tech foreclosure.

20 Some new problem is now problem number 1

30 Get about 70% done fixing the problem (it’s functional but ugly, just need to wrap up this mess so it doesn’t become tech debt…)

40 GOTO 10

I’m a DevOps guy and seem to spend most of my time fixing AI slop. It’s supposed to mean automating builds, tests, scans, deploys, compliance, etc, so the other developers can focus on product code and all the process just works

First of all, there is no graphical stuff. That’s just for simple learning sandboxes.

We have an IDE - Integrated Development Environment. You can think of it as a glorified text editor. We type code in text and it gives us the equivalent of spellcheck, grammar check, autocomplete. They usually colorize the code so you can see structure, match parens and quotes, and other low level assistance. But it gets much more useful with integrations to version control, scanners, build tools, download dependencies . You can click to build, test, scan, commit. They’re usually tons of other tools to make life easier.

But code is cheap and easy to write the first time: much more expensive to fix. Maintenance over time is far more expensive than writing it.

So now we have AI as another tool integrated into IDEs, and it is somewhat useful for generating new code based on patterns from previous code. But it’s never good enough to be an end result. A good developer can use the ai to get a jumpstart on new code, iterate it to get better, and almost always have to use their own knowledge to finish it to a working, maintainable result.

So I have a bunch of junior developers in another country, just directly checking in ai slop. They don’t seem to be experienced enough or diligent enough to recognize when it needs more work. Which means I need to spend a lot more time on code reviews trying to figure out the unorganized mess, give the same feedback over and over, review the same code many times, and inevitably spend much more time on bug fixes for their mess than I would have taken implementing it myself.

The thing is ai is not good at bug fixing. You can try to have it summarize the code, or compare it to best practices but it can’t really help figure out what’s going wrong and how to best fix it. Especially if the original code is ai slop to begin with. So I don’t even get any advantage from it

I… write code. It does stuff. Usually the wrong stuff, until I’ve iterated over it a few times and gotten it to do the right stuff. I don’t “click around in a GUI.” If a tutorial is making you do that, it’s a bad tutorial.

There is this channel that live-streams themselves if you wanna watch like 5-20 minutes and get an overall idea.

https://youtu.be/vc5UPu76XOw

Now it seems every tutorial I see is really just clicking around in a gui. Very little actual typing of code, which is the part I actually find cool and interesting.

Not sure where you’re seeing “just clicking around in a gui”, but if you like computer games, there’s some fun gameplay you can have while coding. Some of those very much contributed to my experience.

BitBurner is a free idle incremental programming game, where you write scripts to hack things to make money to begin with, progressing onto both progressively more complex mechanics (how about automating a manufacturing corporation with a script?) and utility scripts to automate things you’ve been doing manually.

If you like Minecraft, there’s fun to be had with ComputerCraft, scripting things in Lua. With some add-ons (Plethora IIRC) you can access chest inventories via cable and transfer items between them, and set up your own fully automated storage system with recursive autocrafting, as just one example.

Or how about modding games - if there’s a Unity game you enjoy that doesn’t use IL2CPP, like Risk of Rain 2, it’s very moddable using C# and interacting with Unity APIs, and for advanced stuff modifying the underlying IL that C# compiles to. Quite a lot you can learn, and if you stick to pure code mods to begin with, not that hard to get started - though code mod means nothing like new items, new enemies, new characters, buildings etc. since adding models/textures/sounds tends to be more involved.

Something something Jira something scrum agile Confluence something another meeting something hit tab and let copilot do it and repeat.

Disclaimer: I have a dream job for me and my experience is probably not representative.

Go on open.kattis.com, pick a problem, solve it. That’s what 40% of my job is like. 20% more is reading through and understanding where the right place for this bit of code to live or what bits of code I should be reusing to make it. Another 20% is discussing with other engineers the tradeoffs of solving a problem with x vs y and picking what to do, and the last 20% is reviewing code, i.e. making sure other people solve their problems correctly and don’t drop a bunch of hack in our tree.

Mostly I make JIRA tickets.

I think you need to start a project, accept it will be slow and painful, and don’t become an expert before you start, just use the skills you have and see where they take you. The only thing that matters in software is that it works. The definition of working changes over time, but get that first working version and you will keep going.

a full stack developer sufficient in sql and python

Ok, let me first try to explain what happens on a good day, before going cynical.

Let’s assume we have an existing system. You go to what for you appears to be a website, fill some text fields, click on a button, etc. In the background a lot of shit happens. Typically the backend part of the system consists of tens of services each doing it’s own thing. Some participate in returning a response to you, the user. Others just process data further for analytics, security, etc.

One day someone (in most companies a product manager, or a UX researcher) comes up with an idea for a new feature. A user should be able to do XY. And of course pay for it.

That’s where you step in. Since you mentioned full stack, you will need to do everything.

• Create a new page with forms, buttons, nice colors and pictures on the frontend
• Accept the result of user actions of the above to an API in one of the services mentioned
• Save the data into a database (this is where SQL comes into play)
• Retrieve data from a database (SQL again)
• Emit various events or API calls to other services, informing about what just happened

This is all done with code. You can copy/paste, vibe code, just type it yourself. Code is the least of your concern. Making sure it all works together is what’s tricky. You will go through several iterations until you get it right. Then you write automated tests for it (TDD people don’t come at me).

Also you communicate to other people in the company about any dependencies and overlaps with what others are doing. Finally, you can deploy the code to production which will make it available globally to users.

I just described about 50% of the programmer job. I didn’t mention code reviews, architecture discussions, plannings, retros, communities of practice, incident handling, herding cats…

This is all valid in a good case scenario. good company and a good organization in it.
In reality it’s mostly waiting. A lot of waiting. Despair if you can’t make it work. Happiness if you can. Then despair again because all you do is pointless. A lot of fighting against the system designed to make you as unproductive as possible. Or just giving up and faking it for a paycheck.

Here’s the point in which I realized I didn’t understand coding at all.

There as a mod for a game, that was like 95% of what I wanted. It just needed a few values changed, and it would have been perfect for me.

Fortunately, the entire mod was hosted on github, and it was just 3 ‘.js’ files. I figured, hey, how bad could it be? I’ll have a look around, find the values I’m looking for, make a couple tweaks…

All three files were entirely, top to bottom, just INCLUDE statements. Not a single line of ‘code’. And somehow that’s supposed to do something? That’s when I realized I was cooked.

I have never just clicked around in a gui unless its me testing something I’m building. I would suggest finding some better tutorials.

Today I added a few features and fixed a couple bugs.

We have a screen where users can basically click through a list of contacts to send out emails. The list is system generated, and sometimes users want the email to go out to other contacts not on the list. So I added a section to the screen with a typeahead input box to search through all the companies in our system. Once the user selects the company it has to be broken down by branch (region) so I had to build out a small api for getting the company branches and filtering by some criteria to see if they’re even eligible to receive these types of email, so with that data returned I generated a drop down to select the branch, from there another api call to generate a list of checkboxes for each contact at the selected branch. Then I had to build out a mysql table to store the additional contacts to email and work that logic into the code that sends out those types of emails.

Right there you have perl, mysql, html, CSS and JS.

Another thing that everyone’s gonna lose it for when it goes live but took me no more than an hour. A lot of our accounting pages are historical reports, there’s only a few that are updated in realtime, and almost all of them were super slow to load because of all the heavy calculations. I added a couple tables and a cron job to run daily and populate them. So now instead of doing calculations on the fly those pages just pull the data from the db and load instantly.

Sit in meetings