Late last year, I planned to write a post about why I needed to “lock out” more.
I had been spending too much time buried in code. I was constantly at my desk, constantly mentally occupied, and constantly carrying unfinished problems in my head. It started affecting everything else in my life. Even when I was physically present for other responsibilities, my mind was still inside some half-solved bug.
When I code, I can get obsessive. I love solving problems, but that strength can quietly turn into a trap. I was giving coding the best of my energy and leaving everything else with whatever was left.
If you’re a developer, you probably know the feeling.
Between December 2025 when I first considered writing that post and February 2026, when I’m finally putting this down, so much has changed. AI has moved at a pace that’s been hard to track, but incredible to witness.
What once required grinding through syntax and boilerplate can now be delegated. Today, it is possible to outsource the mundane parts of coding to capable, specialized agents and focus instead on intent, direction, architecture, and review.
That shift has saved me an enormous amount of time. More importantly, it has preserved my mental and physical energy.
So what did I actually do?
I setup a team of role-focused agents with clearly defined responsibilities and intentionally selected models based on each agent's role. Their execution is coordinated across product thinking, engineering, and even design decisions.
Here's a quick look at my current agentic team architecture built on OpenClaw
Romeo
│
┌─────────────────────┐
│ Slack Main │
│ Session │
│ (Slack DM) │
└──────────┬──────────┘
│
┌─────────────────────┐
│ Alfred │
│ Orchestrator │
│ (Haiku) │
└──────────┬──────────┘
│
┌─────────────┬─────────────┴─────────────┬─────────────┐
│ │ │ │ │
┌─────────────┐┌─────────────┐┌─────────────┐┌─────────────┐┌─────────────┐
│ Devin ││ Pam ││ Desire ││ Mark ││ Austin │
│ Developer ││ PM ││ Designer ││ Marketer ││ Ops │
│ (Sonnet) ││ (Opus) ││ (Gemini) ││ (Gemini) ││ (Haiku) │
└─────────────┘└─────────────┘└─────────────┘└─────────────┘└─────────────┘
I have been very impressed by the quality of work so far. We are no doubt headed into a future where agents handle all the coding and design work involved in software creation.
I’m aware of the “AI slop” argument. Sometimes it’s valid. But treating that as the full story is a misrepresentation of the state of AI tech. The current generation of coding models can produce strong output quickly, and are improving at a pace that should make every software engineer pay attention.
Yes, there are developers who can still handcraft better code in certain situations.
But nobody can match the speed, iteration volume, and sustained output of a well-orchestrated system of agents, day after day.
And that changes the equation.
In my own network and on social media, I still come across skeptical engineers from time to time and honestly, I understand it.
What’s happening right now feels like identity-level disruption.
For years, engineers were the ones everyone depended on for “the code”. That dependency creates a sense of ownership, pride and control.
Then, almost overnight, a machine starts producing large portions of that same work faster, cleaner, and without fatigue.
Of course people panic.
That reaction is human.
But panic isn’t a strategy.
The practical shift is this:
You are no longer just an executor.
You are now an orchestrator.
You still need deep technical judgment, but your leverage comes from:
Call it creative direction, agentic orchestration or whatever term you prefer. The point is the same: your value is moving upward from typing every line to steering the whole system.
Because these systems still make mistakes.
Sometimes obvious ones. Sometimes subtle ones that only show up under real constraints.
Models are powerful, but they still need:
That is where engineers still matter, a lot.
So no, you are not irrelevant.
But yes, your workflow probably is...if it still assumes 2024 tooling realities.
Since last year, I have repeatedly asked myself:
Is my workflow optimal for the times we are in?
Every time I've asked this, I've discovered something to improve. Sometimes days later, sometimes hours later.
My advice? Adapt your workflow to the tools that are available today as soon as you can. You're leaving a lot of productivity on the table if you don't.
That said, I think a lot about juniors.
I hope they’re finding a way to navigate learning in the AI era — striking the balance between using AI as a multiplier and actually understanding the code it generates, along with the fundamentals beneath it.
If you are still stuck in "this is hype" or "this will replace me so I’ll ignore it", I get the emotion.
But that posture will cost you more than experimentation ever will.
Use the tools and get very good at using them. Push them hard. Learn their limits. Build your review muscle. Upgrade your process continuously.
The engineers who thrive next are the ones who adapt fastest without losing technical judgment.
That’s the game now.