A journal entry, written nine years in. This time not about one sitting, but about the whole journey.

I've been putting this post off for weeks, which is funny, because I'm not shy about publishing feelings on dev.to. I'm shy about publishing the full timeline. Maybe because the full timeline doesn't come out clean. It comes out as a series of accidents I then spent years maintaining.

So here it is, nine years of open source, told in order, with the parts that are embarrassing left in.

2017–2021: The CodeIgniter Years

I joined GitHub on September 27, 2017. The location field on my profile says "Punk Universe", which should tell you how seriously to take the metadata.

The first repos I can see from that era are from October 2017: a minimalist home page, a Laravel 5.5 multi-auth experiment called laravel-multiple-auth. I genuinely don't remember what I was trying to learn at the time. PHP, probably. CodeIgniter definitely. Twig, later. None of those repos matter. They're on the profile because they were the only place I knew I was allowed to put them.

If you open my profile today you'll see 369 public repositories, and if you bother counting forks you'll find that the vast majority of them aren't mine at all. They're other people's projects with my username on top. That's what learning looks like when nobody gives you a syllabus: you fork something, you poke it, you break it, you rarely fix it, you move on. I don't feel embarrassed by that anymore. Until maybe 2021, that was the education.

Some things of mine from those years did get stars — JSONConnection got three, which I remember thinking was proof I was going to be rich in GitHub currency. It was not.

In February 2019, I joined dev.to. I started writing before I knew what I was talking about. Still my preferred order.

None of this era is load-bearing. But everything in it taught me how to read someone else's code before I was any good at writing my own, and that turns out to be the whole skill.

2022–2023: The Punk Phase

2022 was the year I wandered. React hooks. WebSocket clients. An ASCII stickman game. A GitHub profile link generator. I was a freelance PHP developer by day and a JavaScript magpie by night, and honestly I learned a lot and shipped nothing anyone will ever remember.

2023 is when strangers started to notice me a little, in the smallest possible way.

In March I built og-image-rest-generator, a free REST endpoint for generating Open Graph images. Nineteen stars. For me that was the moon landing. It was the first project of mine somebody used without me asking first, and it taught me the difference between writing code and publishing an endpoint — the value was that people could hit a URL and get a thing back, not that the implementation was beautiful (it wasn't).

Around the same time I published The Role of a Software Freestyle Engineer in Modern IT, a "serious imagination journal" where I invented a job title and then spent the rest of the decade answering to it. In December I shipped fck, The fck PHP Framework for Freestylers. Four stars. Possibly zero production users. It was a manifesto in a trench coat pretending to be a framework, and I loved every minute of writing it.

My friends thought "Software Freestyle Engineer" was a joke. Here's the thing about jokes you keep telling: eventually you have to become them.

January–September 2024: The Month It All Changed

I found Turso in January 2024, and I don't say that dramatically. I mean there was a Tuesday where I decided libSQL — this fork of SQLite that runs at the edge — deserved a PHP story, because the whole SQLite world was there for everyone except us.

I started with the thing that didn't need a native extension: turso-client-http, a pure-PHP Hrana-over-HTTP client. First commit: chore: initial commit. Second commit: docs: rewrite readme. Ship it, then explain it — that's the whole strategy, and I still use it.

That one repo pulled the loose thread. A week later I started libsql-php-experimental. In April, libsql-client-php — an FFI-based client for native libSQL. In June, the native extension, the Laravel driver, and the Doctrine DBAL driver all became first commits in the tursodatabase org, under upstream, with messages like chore: ohmypunk and chore: idk what i am doing but i love it. I stand by those messages. They were honest.

On June 19, 2024 I shipped turso-php-installer, because I wanted the extension installed, not because I wanted a Packagist package. First commit: chore: i just want it. It has 22 tagged releases now and about 10,750 total installs, roughly 1,100 a month. August brought the Platform API client. September brought the Laravel HTTP wrapper, then the ecosystem repo, because at that point I had more repos than I could hold in my own head and someone needed a map. Probably everyone needed it, but me first.

By the end of 2024 I had accidentally built the thing I used to wish existed: a coherent Turso PHP ecosystem — extension, installer, HTTP client, Laravel driver, Doctrine DBAL, Platform API SDK, docs. From a freelance desk in Tegal. Between paid jobs.

Nobody paid me for any of it. I want to flag that now, because it comes back later.

February 2025: The Lone Wolf Hangover

In mid-February I shipped Whatsback Web — a free, self-hostable WhatsApp provider. It has 25 stars. What it didn't come with was the thing I'd quietly expected after sharing it: a wave of other people showing up to build with me.

I wrote a thread of posts that month. Did you know about this?, the original case for it, It's Free and Open Source!, and then, honestly the worst one, what did I get in return.

The answer was: more visibility than reciprocity. Views without a single maintainer stepping up. It ended with From Fun to Frustration: A Lone Wolf in the Open Source World, which is the loneliest post I've ever published and the one people still quietly DM me about, years later.

I wasn't looking for sympathy. I wanted company. There's a difference. One of those asks for attention; the other asks for help, and help requires people to know you need it. If you're building in the open and wondering why nobody shows up: sometimes it's the mechanism, sometimes it's just the quiet, and sometimes it's that you made something free and then discovered free is not the same thing as loved.

Most journey posts edit this part out. I'm leaving it in, because if I cleaned this timeline up, the rest of it wouldn't make sense.

March 2025 to Today: Becoming Someone Who Maintains Things

In March 2025 I shipped mylibsqladmin, a modern web UI for libSQL servers. It's my most-starred original repo at 58 stars, which tells you everything about how GitHub counts popularity. What it taught me is the difference between shipping and sustaining. Shipping is one weekend. Sustaining is every commit after that: the v1.5.6 tag went out on November 16, and I spent that weekend in verification-screenshot hell making sure the database management flow never deletes your data on archive. The commit message says it better than I can: never lose data on archive/restore/force-delete; honest toasts.

October 2025: escpos-printer-server. Twenty-seven stars. Everything I touch eventually needs to print receipts; I've made peace with this. December: ffrs — ffmpeg but with plain English, my little act of rebellion against command lines. August 2026: duitku-ts, a TypeScript SDK for Duitku, an Indonesian payment gateway. Zero stars so far. It's fine. Some repos exist for the users who need them, not for the star counter.

In September 2026 I wrote AI Can Write the Code. So What Are You Actually Good At?. The short version is that when writing code becomes cheap, understanding it becomes expensive, and no model is going to sit in the release room with you deciding whether the delete envelope really comes back as a scalar string. That's a maintainer's call, and it's made of experience, not tokens.

And then November 2026 happened, all at once, the way it does.

On November 9 I merged PR #27 into the extension: a reliability overhaul — FFI error mapping, a dangling pointer/double-free fix, lazy streaming actually streaming. I published the extension as v1.7.0. The installer finally closed issue #10 from March (2.2.1), and absorbed brendt's PR #11 (2.2.2). The Laravel driver shipped 1.4.0 and is now at 1.7.0 with WilliamIsted's PR #114 — the fix that turned silently-swallowed constraint violations into actual, catchable exceptions. I described all of it in Maintaining the Turso PHP Ecosystem in My Spare Time.

Three days later came The Sitting of November 12–13, 2026: turso-api-client-php v2.1.0 — 53 of 53 OpenAPI operations, builder-only, and a delete path that had been TypeError-ing in production-shaped data — plus turso-platform-laravel v1.0.0 and v1.1.0, where PHPStan level 6 caught getenv() returning false and the ?? '' that never fired. Two blank release notes. One red CI. A scratch database that was created, filled, read back, and deleted as proof.

Nobody tells you this part: verification is the job. Shipping is the glamorous half. The other half is standing in front of CI like it's a hostile witness until it admits what happened.

The Numbers, Because You Probably Want Them

  • 369 public repos. 73 original. The other ~296 are forks I used as training wheels and mostly never returned.
  • 313 stars across my originals. Best-decorated: mylibsqladmin, 58.
  • ~10,750 Packagist installs of the installer. ~1,100 a month.
  • Money from open source: zero. Not a cent. My READMEs have GitHub Sponsors and Saweria links that behave like honor boxes, which is to say they are there, and mostly they are ignored, and that's allowed.
  • Paydays that kept me going: the tursodatabase org letting me hold the PHP side of the house, brendt merging a fix I didn't write, WilliamIsted fixing a data-integrity bug I hadn't seen, and the quiet DM from someone saying the docs finally made sense.
  • The ecosystem README put it better than I can: People call me "crazy" and that's true. Because I work Full-time for Open Source.

It isn't a profitable ledger. It also isn't a loss. It's just... mine.

What Open Source Taught Me (The Short Answers)

Shipping beats planning; everyone I respect ships the ugly version and lets it breathe. Trust evidence, not résumés — the scratch database being deleted is the receipt, and anything else is vibes. Open source is maintenance, not invention, and the ecosystem doesn't need another package as much as it needs PR reviewers. Loneliness is the ambient condition, but so is company — the fixes I didn't write are the proof that it isn't actually just me in there. And AI hasn't changed any of this; it writes plausible diffs, but someone still owns whether the delete envelope comes back as a string. That someone is the maintainer.

The rest I'm still learning, probably on a Tuesday at 1 AM with CI red.

What I Need Right Now

  • Review turso-client-php PR #26 — ARM64 Linux builds and the LIBSQL_NUM work. It gates the Doctrine DBAL 4 story.
  • Weigh in on turso-doctrine-dbal PR #1 — the DBAL 4 path.
  • Windows PHP 8.5 NTS binaries for the extension, aimed at v1.7.1.
  • Honest maintenance status for turso-client-http — say so if it's not actively maintained, or fix it and say so then.
  • Docs review. Find a wrong claim, file an issue, be that person.

You don't need to build a new framework. The ecosystem doesn't need another repository for the sake of repositories. It needs reviewers, testers, and people who report bugs they can reproduce.

Nine years in, from a forked CodeIgniter boilerplate in October 2017 to a PHP ecosystem for Turso that didn't exist before I decided it should — that's the whole arc, ugly and ongoing. The work is the receipt, and the work also needs reviews.

Same ask as always. Be one of them.

Previous posts in this series: Maintaining the Turso PHP Ecosystem in My Spare Time · The Sitting of November 12–13, 2026 · From Fun to Frustration

Support the work: GitHub Sponsors · Saweria · X/Twitter: @panggilmeiam

License: CC-BY-SA. Code in the linked repositories; errors live here because I wrote them.