You Will Know Nothing And Be Happy
It’s 2030, and your boss just asked you to pull data to help better segment your users and understand their behaviors.
You open your laptop, and three agents spin up immediately
You provide the general ask, and they start parsing your information schema, past query history, previous analysis, as well as current trends and market behaviors from public databases.
Python scripts are being written.
Queries run.
Do you know how any of it works? No.
Do you know the difference between a right and a left join? No.
But those are meaningless details.
Thousands of lines of code, and who knows how many SnowBrick credits later, you’ve built a customer segment table. Do you run any form of QA check? Nah, it’s probably right.
Plus, you’re busy scrolling through your AI-generated TikToks and probably couldn’t focus long enough to check.
You will know nothing and be happy.
Do You Even Know What Your Code Does?
One of the skills you pick up as an engineer is your ability to hold multiple functions, entities, abstractions and workflows in your head.
You start to create mental maps of your code base, I imagine not dissimilar to those our parents had when they used to give directions with street names and landmarks.
Those mental maps of your code allow you to easily start to trace how changing one function impacts another and how even a small change could impact an entity or dependency several functions away.
It makes it easy to debug when there is an SEV because you likely have several well-founded theories of what caused the issue based on looking through the past few PRs.
But hey, why remember anything?
Why remember how to describe the world around you? The GPS will do that for you.
Why remember how your code base, that you haven’t truly looked at for 6 months, looks like? The AI will do it for you….
If that’s the case, then it’s hard to know what the code changes I am making are really doing, and it’ll be even harder to debug on a larger project.
It looks like Amazon may have some reasons why it’s valuable.
Just giving mid and junior engineers access to a magic box that can confidently produce code that may or may not be the correct code, but looks like it works, will eventually have most of us turn off our brains and just trust the code works.
We forget how to think even a few functions deep.
Just rewrite the whole set of functions!
Because reading other people’s code is not exactly a fun activity.
So why not just rewrite it?
That’s probably why we often see the meme below in multiple variations. It’s because when most of us see a long PR, our eyes glaze over.
And the code changes AI can make can be far more then a few hundred lines.
The challenge then becomes, can you understand the implications of every change?
Do you even try?
Or do you just run it through a few tests and say, yeah thats good?
The Illusion of Understanding
I think we are currently going through the stage where many of us are just flipping to the end of the math book.
Looking at the solution and assuming we understand our code bases, our systems, and other complex workflows that live in experts’ heads.
After all, the automation we just built works, right?
There is no need to slow down and actually understand the problem being stated. Why?
You’ve got 10 agents running, and they need new tasks.
It’s unproductive to slow down and ask questions.
Just iterate again; it costs you nothing. So maybe this next roll of the dice will get you the right output.
And when there is a bug, don’t worry, just throw it into Claude or ask your pilot to fix it.
All the while, you build no understanding of what is going on in your code base. Your lines of code grow, and the system continues to be built like a Jenga tower waiting to crash down at some point in the future.
Final Thoughts
The problem I see is that we are going from a world where there was at least one John or Jane at a company that understood the systems they built, to not even that.
And you can already see hints of this today.
More SEVs
Longer debugging cycles
Increased reliance on “just try something” fixes
Seniors being pulled in to validate what juniors (or AI) shipped
I want to be clear. I am not over here saying you shouldn’t use AI. It’s been a great tool, and I have found plenty of ways it can help improve my workflows and help create tools that can make my life easier.
But there’s a difference between:
Using AI to accelerate your thinking and using AI to replace your thinking!
We are all figuring out where these tools will play an impactful role in the future, but if the only skill you’re picking up on that path is copy-pasting the output from your LLM, you’re going to have a bad time.
With that, I want to say, thanks for reading!
Are You Looking To Improve Your Data Infrastructure?
Todays article. I wanted to let y’all know that today’s article is sponsored by me, the Seattle Data Guy!
Our team has helped dozens of companies turn data in to actual business outcomes. We’ve also helped companies set-up their data stack from the ground up as well as untangle their current data infrastructure. If you’re looking for an experienced data consulting team to help you set-up your data infrastructure and strategy, then set-up some time with me today!
Articles Worth Reading
There are thousands of new articles posted daily all over the web! I have spent a lot of time sifting through some of these articles as well as TechCrunch and companies tech blog and wanted to share some of my favorites!
Why Your Data Stack Won’t Last - And How To Build Data Infrastructure That Will
As a consultant, I have been called in to review and, in many cases, replace dozens of half-finished, abandoned, and sometimes forgotten data infrastructure projects.
The data infrastructure in a few cases may just need a little tweaking to operate effectively, but other times the project is either so incomplete or so lacking in a central design that the best thing to do is replace the old system.
Trust me, I’d love it if I could come into a project and simply change a few lines of code, and then everything would just work. However, so many projects are filled with unclear design decisions or resume-driven development that were never rooted in good planning.
Of course, business stakeholders may have also push to get things done quickly. Forcing data teams to take on tech debt that will never be fixed. Don’t get me wrong, you want to get things done and move projects forward. But taking on technical debt is a decision that needs to be made intentionally. Otherwise, like in resume driven development, your data infrastructure might disappear.
This begs the question.
Issue #47 – The Misjudged (Yet Integral) Role of Data Governance
If there is one data domain that organisations undervalue, it is Data Governance.
I mean, nobody is truly enamoured with the idea of government and bureaucracy. You receive complaints about it being inefficient, creating red tape, and taxing you for your hard-earned money, among other issues.
So why add governance into our corporate structure?
Does the data industry really need it?
I mean, especially since Data & AI has a future-forward and cutting-edge reputation, how can we build fast and break things with governance? We need to invest to stay ahead of the curve and get started immediately without delay!
And governance gives off the opposite reputation: it’s a bureaucratic barrier to progress; it is required for risk purposes; it’s hard to execute against and do properly.
End Of Day 214
Thanks for checking out our community. We put out 4-5 Newsletters a month discussing data, tech, and start-ups.
If you enjoyed it, consider liking, sharing and helping this newsletter grow.
Good point on code but data is actually scarier. A broken function throws an error. A broken segment table just looks fine. Wrong assumptions quietly bake into decisions for months before anyone notices.
Great article. What I think techie people tend to not fully understand or internalize is that no one outside of about 3% of the population gives a shit about how the code works. No one. They want the end result. You’re stuck thinking that your craft matters or that there’s an inherent nobility in understanding and modeling complex systems but until you can show how that knowledge is still necessary to drive the result you will lose out to AI.