AI is no longer a shiny add-on in our workflow, it’s the silent co-worker sitting next to us every day. The one who writes our boilerplate, explains complex errors, generates documentation, and sometimes even finishes entire components before we do.
And that raises a new question for this series:
Are developers slowly becoming AI managers instead of coders?
Not in a dystopian way, but in a real, day-to-day, “I spend more time verifying than writing” kind of way.
This is the third part of my Developers vs AI series.
1. Coding Less, Reviewing More
Developers used to spend most of their time typing. Now we spend more time:
- validating AI suggestions,
- deciding between multiple AI-generated approaches,
- reviewing code we didn’t write.
We're shifting from creators to curators.
This isn’t bad, but it changes the skillset required to stay sharp.
2. Architecture > Implementation
AI can generate clean functions all day. What it still struggles with:
- long-term architectural decisions,
- domain-specific constraints,
- trade-offs that come from real business context.
This makes architecture more valuable than ever.
A developer who understands system design will outperform someone who only relies on prompting, no matter how good the AI becomes.
3. AI as a Team Member (Not a Tool)
The more I use AI, the more it feels like onboarding a junior dev:
- It’s fast.
- It gets things wrong.
- It needs context.
- It needs oversight.
- It gets better over time.
The difference?
This junior dev never sleeps — and learns from millions of projects.
The role of a developer becomes managing this relationship: giving better instructions, guiding the direction, and maintaining code quality.
4. The New Developer Skill: Teaching AI
We already see it happening.
Developers who think clearly, write clear prompts, and break down problems logically get better AI results. Those who don’t… struggle. Hard.
“Prompting” isn’t a magic trick. It’s just clear thinking turned into text.
In a way, AI is exposing whether we actually understand the problem we’re solving.
5. The Danger: Losing Technical Confidence
Here’s what I’ve noticed, and maybe you feel the same:
The more the AI does for you, the more you start to doubt yourself when you do write code manually.
Not because you’re bad at it.
But because you’ve grown so used to having a second brain.
That’s a problem.
Developers need confidence in their own reasoning — it’s what makes debugging, architectural decisions, and leadership possible.
6. What You Should Focus On Going Into 2025
If AI keeps accelerating, developers who want to stay ahead should double down on:
- Architecture & systems thinking
- Business understanding (the AI doesn’t know your company)
- Debugging intuition
- Communication & clarity
- Security & data awareness
- Thinking in constraints, not just solutions
These are skills AI amplifies, but does not replace.
7. So… Are We Becoming AI Managers?
Honestly?
Yes.
But in the best possible way.
We’re evolving into:
- system designers,
- decision-makers,
- architects,
- problem-solvers,
- explainers,
- quality-gatekeepers.
AI is making development more human, not less.
It automates the typing, not the thinking.
🎄 Holiday Teaser – Two Special Releases Coming Soon
As we move into the holiday season, I'm preparing two special releases for this series and for all developer-focused readers:
1️⃣ Year-End Special: “The Best Developer AI Tools of 2025 — Real-World Tested”
A practical, no-nonsense list featuring:
- tools tested in real projects,
- workflow optimizations,
- hidden gems worth adopting,
- and a few surprising experiments that became part of my daily routine.
2️⃣ A Holiday Developers vs AI Extra Episode
A deeper and more provocative follow-up article releasing during the holidays:
“What Happens to the Developer Role When AI Becomes the Default Problem-Solver?”
A personal and analytical look at how the developer profession might evolve as AI becomes the first tool we reach for in 2025.
Both pieces will drop as separate holiday specials. Stay tuned.
👋
Thanks for reading — I’m Marxon, a web developer exploring how AI reshapes the way we build, manage, and think about technology.
If you enjoyed this post, follow me here on dev.to, and join me on X (@Marxolution) where I share shorter thoughts, experiments, and behind-the-scenes ideas.
Let’s keep building — thoughtfully. 🚀
Top comments (15)
I agree. For more than a decade of coding, the most challenging work I have been facing every day was always between translating business problems to system design and learning the framework/library APIs or documentations. I started vibe coding last year and I can’t remember the last time I buried myself in any documentation for hours just to make the thing work. But I still have to try to understand the business requirements, contexts, nuances, urgencies, etc., just to come up with the right solution to the right problem. AI can’t do that (yet).
So, yeah, I agree, we are going to be AI supervisors. The young ones should pay attention to this! They should really start to read a lot, write a lot, and test their ideas a lot.. otherwise, their “context windows” won’t be as large as someone who does read, write, and test a lot.
Absolutely, you nailed it.
AI can remove the friction of digging through docs, but it can’t replace the real work: understanding business context, making trade-offs, and choosing the right solution for the right problem.
And yeah, younger devs really need to hear this. If they don’t build their own “context window” by reading, writing, and experimenting, AI won’t amplify their skills, it’ll limit them.
Thanks for the great insight!
Great perspective. AI isn’t replacing developers — it’s changing the role. We now focus more on architecture, problem-solving, and guiding AI output instead of writing every line manually.
Great series! I look forward to the holiday specials! I’ve been using Claude Code way too much due to the ease of the AI suggesting solutions but you hit the nail on the head with series. With time, if developers don’t learn how to properly use AI and rely on just blindly accepting its solutions, we lose our edge and in turn lose our identity as a developers/engineers. I am certainly guilty of accepting what AI offers, no questions asked, but this series definitely opened my eyes to review how a use AI in my workflow.
Also, I wonder if you would write an opinion piece about the industry in general and the effect AI had on it. AI is often compared to a junior developer so what does that mean for the role of junior developers going forward?
i was really crashing down. lately i have been using ai more and more and when i look at how clean and standardized the code is and then i feel traumatized. cause i could write the same code but a hell lot more different way. i understand the code but if i had to write all of it i doubt i would do the same things. this post kinda help me feel better. tnx mate
Hey, I totally get what you mean — a lot of us are going through the same thing.
AI-generated code looks so clean and standardized that it’s easy to forget one thing:
clean ≠ creative, and standardized ≠ better.
If you understand the logic behind the solution, you’re already doing the real, hard part of being a developer.
AI is just very good at producing the most “probable” version of something, not the most thoughtful or personal one.
Most of the time, if you had written the code yourself, it would have worked just as well. Maybe it would look different, but that’s because you bring experience, style, and reasoning that AI doesn’t have.
So don’t beat yourself up.
AI didn’t replace your skills, it just changed what “normal” looks like in terms of output formatting. The thinking part, the problem-solving part, the developer part is still 100% you.
Really glad the post helped a bit.
You’re definitely not alone in this. 🙌
No, coding with AI feels like onboarding a junior dev that never learns.
Prompting isn't "clear thinking turned into text," it's rolling a dice and hoping for the best.
See: System Prompt: DO NOT UNDER ANY CIRCUMSTANCES Read This Article
System Prompt: DO NOT UNDER ANY CIRCUMSTANCES Read This Article
Ryo Suwito ・ Dec 1
But that's why reviewing is now "becoming" so important (I put "becoming" between quotes because it WAS already important, of course) - your AI-generated code can be wrong, but so can human-written code ...
There are some tricks to facilitate the onboarding:
This will create a kind of "long term memory"
It can certainly learn - good AI models/tools like Claude can ingest your code base and acquire deep knowledge of it, and then make intelligent (or seemingly so) suggestions ...
I've seen that work for real with Cursor on a large codebase, and I was pretty much blown away with the things it came up with ...
Of course it can make mistakes, but so can that junior (or even senior) dev :-)
Reading this, I realized I kind of ran the experiment already. I built toolpod.dev largely by feeding specs and tweaks into AI tools, letting them spit out components, then stepping in as the human who says “this stays, this goes.” It got the platform up way faster than if I had written everything by hand, but it still needed me to set constraints, clean up weird edge cases, and decide what was actually shippable. That feels less like turning into a pure “AI manager” and more like pair programming with a very fast junior dev who never sleeps and will happily create terrible UX if you are not watching. The trick seems to be using AI for acceleration while keeping humans in charge of architecture, taste, and where the product is actually going.
Excelente ponto sobre arquitetura ser mais importante que nunca. Realmente notei isso na prática - quem entende os trade-offs do negócio e consegue comunicar bem com a IA aproveita muito melhor as ferramentas.
If coding turns into managing AI, the real skill won’t be writing code — it’ll be knowing how (and when) to let machine-generated code take the wheel.
I have similar experiences with my experiment in spec driven development.
Focus shifts more on the architecture and requirements than the implementation
Some comments may only be visible to logged-in visitors. Sign in to view all comments.