I do not read Hacker News

It is New Year’s Eve in Tokyo, which means that I’m in a cafe, my Notion vision board out in front of me, looking at the dozen plus 2025 resolutions I’d jotted down in January. I wanted to read 52 books this year, to bike the Icefields Parkway, to learn to French Braid my hair (which is harder than you think to do well), to host 4 events. Etc, etc, etc. I go down the list: marking most incomplete (25/30 ski days on the mountain, 32/50 books), some done, and deferring a few for next year.

Eventually I arrive at one, more serious than the others and near the end of the list: “gain career clarity,” the card says. Funny how three words can so casually capture three months of therapist-mediated angst.

The checkbox looks back at me blankly.

My cursor hovers over it, motionless.

For a good part of 2025, I’d been working on a project at Nextdoor that in theory should have fascinated me. And in many ways, it did, but only in the way a good crossword might hold your fascination for the duration of a morning coffee. A quiet satisfaction that lasts until breakfast is over; you start the day; tomorrow morning arrives, and another NY Times crossword is in your inbox, yesterday’s already well forgotten.

This isn’t to say work needs to be some crusade you take up and make into a life mission. But this year, I realized I’d begun to atrophy, going “smooth-brained” if you will.

I had gone from someone who would eagerly chase a particularly tantalizing thread of work through lunch and dinner and Thanksgiving break to dragging my feet through the work day. I could barely recall that I’d once been someone who’d made every career decision with growth in mind: from taking a college gap year to work full time at Apollo, to passing on a safe new-grad bet at Meta to follow some of the best engineers I’d met to Nextdoor, to joining a new team to rebuild our ads platform and reporting pipelines from the ground up, and then switching teams again to go deep on product infra I hadn’t touched before: tiered caching and optimizing ORM database access.

By the time Q4 of this year came around, I was coming to terms with the fact that a change would soon be in order.

Still, when Walt (whom I’d worked with at Nextdoor) floated the idea of working together again at Natural, I dismissed the thought immediately.

In my head, there was and, perhaps still is, a fuzzy mental picture of what a start up engineer should look like, and it isn’t me. It’s some amalgamation of forked repositories on Github and Saturday-evening side-project-hacking and technical microblogs still followed via RSS. It’s an engineer who reads Hacker News, and has open-sourced a personal repo to deal with some niche, dev inefficiency years before an official solution exists. It’s someone who has Twitter (yes I still call it Twitter) but is probably also on Bluesky now. And more importantly, it’s someone who puts together one vibe-coded demo after another to help the company navigate the elusive path to “product market fit.”

I, on the contrary, could write love letters to sealed classes and Kotlin when clauses. I could wax poetic on compile-time exhaustiveness safety for enums and perform sonnets on init-time validation in immutable data classes. And I would happily do a miniature Ted Talk on the value of type-level null-safety. (So sultry of a concept, I consider it borderline lewd.) You’d be hard-pressed to find me duct-taping pipes just to “make something work” for a customer call, and AI slop makes me nauseous.

“I don’t want to [insert start up spiral of throwaway eng work and feature churn]”

“You won’t need to,” Walt says. “You’ll be working on money movement.”

Nov 1st, 2025 is turning into an awful night for Toronto Blue Jays fans, courtesy of LA Dodger Yoshinobu Yamamoto who has been sent back to the pitcher’s mound, less than 24 hours after pitching the night before. (This is something typically unheard of, given that pitchers tend to stick to a rigid cycle of one day on, four days rest, for full recovery). I know because World Series Game 7 is playing behind me at the office, while I work on account transfers in our ledger. It’s day three of my work trial at Natural, and I’m tracing a full payment lifecycle—from authorization to settlement to reversal—through our system, unaware that the night is slowly creeping past 8 then 9 and then 10.

At some point, Eric forces me to go home, which I do, reluctantly. My fingers are itching, my mouth salivating. I think about the ledger transfer invariant validations I’m going to try to make airtight tomorrow.

When I get home, there’s my bed. And there’s my laptop.

Maybe just another hour?

I accidentally work until 2am that night.

Building a ledger, a payments system, is largely a solved problem. Many other companies have done it before (to varying degrees of correctness). In fact, it’s one of the more straightforward problems Natural will need to solve if we’re evaluating in terms of technical or product ambiguity. And it’s merely one of the first, foundational items Natural needs to check off before getting to the heart of differentiating product.

But for some reason, working on this has me nerdsniped to the nth degree. Late November, when I told one of my mentors I was thinking of joining Natural, they ask me, why not somewhere like [redacted AI start up] instead, where more novel challenges lay.

“Don’t you want to scale GPUs for the world’s largest neural network in the world?” he asked.

I squinted and shrugged. I, in fact, do not want to scale GPUs for the world’s largest neural network in the world.

Perhaps, it’s because payments, at the end of the day, plays directly to my most pathological engineering desires. It feeds the part of me that is a sucker for correctness—the part of me obsessed with Kotlin when clauses, that wants every state enumerated, every edge case overturned, sharp flint polished over and over until not a single jagged edge remains.

So when I think about the system I’m building, at replaying events, at complete traceability, I go to bed hungry and I wake up ravenous.

Me being at Natural isn’t something so deep as a “life mission” (sorry, Devan). And I can’t talk at length about the product vision for the future of payments yet (I’ll direct you to Kahlil for that, but give me a few months and maybe?). In fact, agentic payments is still a space I’m wrapping my head around product wise. Right now, I just want to pick the brain of every engineer with payments infra experience I can talk to. I want to ask about event versioning and invalid states and oh, how did you handle reconciling out-of-order events from third parties for internal correctness?

In full honesty, I’m also still learning the ropes, figuring out how to rein myself in before I burn out, how to juggle the urgency of a start up, against my inherent bias towards cautious solutions, and the strict demands for correctness that comes with systems moving real money.

For me, the energy shifts from day to day. Some, it’s easy flow: long hours of pure focus, natural stretches of grinding in the office; others, there’s anxiety: the low-grade awareness of how much there is to build, of time ticking by. When the anxiety creeps in, I chat with Eric, and we rehash the game plan, how we’re going to balance speed of execution, what needs to come first and what to shelve for now, how we’ll sequence all the pieces that will eventually slot in.

It eventually dawns on me that I’m being pushed in ways I hadn’t expected; and I can feel myself shifting to meet the work. Two days after Christmas, mid-holiday on the other side of the world, I find myself scouring through repositories of well-known open-source workflow engines while jostling about on Shanghai Metro Line No 11. For the first time in years, I’m scrolling through Hacker News, where I eventually discover a random Substack article and rabbithole into some of Claude Code’s more obscure features. There’s so much to do that it’s been on my mind to sit down and figure out where I can scale myself further.

Maybe there’s a hint of a start up engineer inside me.

Back in Tokyo, I’m sitting in the cafe, the vision board still open, the checkbox still empty. The ice in my matcha has long since melted.

Those three words — gain career clarity — don’t carry the same weight they did a year ago. I still don’t have answers for three years from now, or five, or ten. In fact, I don’t know if I’ll still be a software engineer then, or what that title will even mean. But I do know what I want to be working on right now—and who I want to be working on it with. For now, that seems like enough.

Join us /careers.