A couple of weeks ago, I decided I should probably just build my own email client.
Not because I’m a developer who lives for that kind of thing. Not because every other option had failed me. But because I wanted email tracking without using email in the browser, and my brain, primed by a year of AI-assisted building, looked at that problem and thought: “Yeah, I could fix that.”
That’s the world we’re living in now. And honestly? It’s both incredible and a little bit terrifying.
The Itch That Started It All
Let me back up.
I use Google Workspace for email, and one thing that matters to me is email tracking, knowing when someone opens a message and when/if they click links, so I can time follow-ups, gauge interest, and hold myself accountable to conversations.
The best way I’ve found to do this is with a browser extension called Mailsuite (formerly Mailtrack). Simple, effective, works great.
Except it only works in the browser.
And I genuinely hate managing email in the browser. It’s not a dealbreaker on its own, but it’s cognitively expensive in a way that compounds over time.
I use Arc, which has spaces, so opening an email means hunting down the right tab in the right space while something else is demanding my attention.
The browser is a powerful environment, but it’s also noisy. I’d much rather work in a dedicated app, where switching contexts feels intentional rather than chaotic.
So the itch was simple: I want email tracking without the browser.
In the past, I would have accepted that tradeoff. I’d have grumbled, searched for a workaround, maybe found a native mail app that almost did what I wanted, and eventually shrugged and moved on
The gap between “I wish this existed” and “I could build this” was wide enough that the idea would stay safely theoretical.
But we have AI now.
So instead of shrugging, my brain immediately went to: “Why don’t I just build an email client that lets me use the extension?” One thought led to another. I started researching browser wrappers. I looked into existing apps that might already solve this. I started mentally scoping out what a build would look like.
Before I knew it, I was three rabbit holes deep into a problem I didn’t have yesterday, working on a solution to something that wasn’t even close to my most important priority.
That’s the trap. And once you see it, you start seeing it everywhere.
When the Filter Breaks Down
For most of my life, there was a natural pacing mechanism built into problem-solving. Friction.
If I had an idea, I had to weigh whether it was worth the cost.
Would it take hours? Days? Weeks?
Did I have the expertise?
That friction wasn’t always fun, but it acted like a filter. It forced prioritization by making the cost of a detour real.
AI doesn’t eliminate that cost. It just makes it invisible.
When I get the itch now, the distance between “wish” and “action” has collapsed to almost nothing. I can scaffold a concept in an afternoon.
I can prove something is viable by the end of the day.
I can have a working prototype before I’ve fully thought through whether I actually need one.
The capability is real, the speed is real, and the excitement that comes with it is absolutely real.
The bill comes due later.
Every project I start is a project I have to maintain, revisit, or abandon. Every detour is time borrowed from something else. Every prototype that doesn’t get finished lingers in the back of my mind.
The Pile
Over the last year, I’ve probably had more than 20 ideas turn into real builds. Add-ons, scripts, plugins, workflows, experiments, and tools to support other tools.
Some of them got finished. Many didn’t. A handful are stuck in that uncomfortable middle state where they work well enough that I can’t delete them, but not well enough that I can realease them.
On the surface, that sounds like output. Look at all the things I made.
But when I step back, I see a different pattern. I see a pile. I see attention fragmented across too many things. I see the feeling of constantly being mid-sentence, never quite landing, never quite done.
Productivity has a texture, and this doesn’t feel like productivity. It feels like spinning.
The clearest example came after finishing a large project where I built a whole suite of custom plugins and integrations.
During the build, I tried to stay organized because the complexity demanded it. Then the project wrapped. And now, cleaning up, I’m finding stuff everywhere.
Notes without context. Repos that made sense in the moment but don’t have a clear home now. Half-finished experiments that are no longer attached to the problem they were supposed to solve.
This is what sneaks up on you when you build quickly with AI: your ability to create outpaces your system’s ability to absorb it.
You can generate a mess at the same speed you generate solutions. And the mess is quieter, so you don’t notice it building until it’s already everywhere.
The Justification Machine
What makes this particularly hard to fight is that the detours always feel justified in the moment.
“It’ll only take a day.”
“I can prototype it quickly and decide if it’s worth pursuing.”
“This friction is costing me more time than the fix will.”
“I’ll just see if it’s even possible.”
Those statements are often true. That’s what makes them dangerous.
They account for the cost of building but not the cost of the distraction. They account for what gets created but not what gets left behind.
They make a single detour sound reasonable while quietly ignoring that you’ve taken twelve of them this month 🤣.
AI makes bad ideas feel like good ones. It makes genuinely good ideas feel urgent.
Ideas that were always worth doing someday now feel doable today, and “someday” becomes “right now” before you’ve decided whether right now is the right time.
More Capability Means More Discipline, Not Less
Here’s the counterintuitive thing I keep running into: having more capability hasn’t reduced the need for discipline. It’s increased it.
When you can’t do everything, prioritization is built into the constraints. When you can do almost anything, prioritization becomes a skill you have to actively cultivate.
The absence of limits doesn’t set you free. It just means the limits you need have to come from somewhere else, and that somewhere else is you.
AI a power tool.
A table saw makes a skilled carpenter faster. But it also makes mistakes faster. It doesn’t replace the need for precision and planning.
If anything, it raises the stakes for having both, because the speed at which things can go wrong scales with the speed at which things can go right.
The same is true here.
AI gives me the ability to build more, build faster, and build things that would have taken weeks in an afternoon. That’s genuinely valuable.
But it doesn’t make focus less important.
It makes focus more important. Because the cost of being unfocused has gotten higher right alongside the cost of building the wrong thing.
What I’m Actually Working On
The answer I keep arriving at isn’t “stop building things” or “AI is distracting.” It’s simpler and harder than that: I need a better operating system for how I work.
The piece I was missing isn’t capability.
It’s a reliable way to separate interesting from important.
A system that captures ideas without immediately turning them into projects.
A way to protect deep work from opportunistic detours, not because the detours are always bad, but because I want to choose them intentionally instead of stumbling into them.
Part of that is structural.
I’m getting better at keeping a running list of ideas that’s reviewed on a schedule rather than acted on the moment they arrive. I’m building in a delay between “I should build this” and “I’m going to build this.”
I’m trying to ask one question before starting anything new:
What is the actual problem here, and is building something truly the best way to solve it?
Sometimes the answer is yes.
But sometimes the honest answer is that I want to build it because building things is fun and I’ve got a good excuse.
The other part is accepting that finishing something has value that starting something doesn’t. Every completed project is a thing I can actually use, actually reference, actually point to.
Every abandoned prototype is just clutter, however interesting it may have been to build.
Moving Forward
I want to be clear that I’m not down on any of this.
I love what AI-assisted building makes possible.
The feeling of sitting down with a problem and actually solving it, without needing to wait for a developer, a budget, or a six-week timeline, is something I don’t take for granted.
That capability is real, and it’s changed the way I work in genuinely meaningful ways.
But I’m also done pretending the tradeoffs aren’t real.
What I’m dealing with is shiny object syndrome, upgraded.
Same impulse, same fractured attention, same pile of half-finished things. Just operating at AI speed, which means the pile grows faster and the distraction arrives dressed as productivity.
The antidote isn’t less ambition. It’s better guardrails. Not guardrails imposed from the outside, but the kind you build for yourself once you’ve seen clearly what the problem actually is.
I’ve taken some time to start systematizing my work and building workflows and tools that reduce the number of moving parts. Looking forward to sharing this with you in a later post.
