Software-Dev on Andrew Quinn's TILs

The long death of programming languages

Wed, 25 Feb 2026 00:00:00 +0000

I wrote my first Clojure program in anger a few months ago. Only, it’s not accurate to say that I wrote it; AI did most of the writing, and I directed it on what my design goals were and why. Architecturally, you could call it was a stateless anti-corruption layer, yapping between some REST endpoints and some ancient proprietary serialization.

I had never written a Clojure program of nontrivial size before. Indeed I had barely touched any Lisp at all since working through SICP during a high school summer vacation, over a decade ago. That didn’t matter. Much more important was emphasizing certain invariants from the start, like the aforementioned statelessness, which made it very easy to unit and integration test, which was the true challenge.

Lua is a pretty good config language

Tue, 06 Jan 2026 00:00:00 +0000

Lua and Scheme occupy curiously similar places in my noggin, becuase my primary use case for both has been as scripting languages embedded into other, larger projects. Lua of course is basically designed for this job, but it’s a solid intermediate format to target even if you have no immediate intentions to use the programming language parts at all. A TSV file like

1
2

10.0.0.5  192.168.1.1 80  tcp
10.0.0.6  192.168.1l.1  443 tcp

can be given a lot more structure on the cheap by writing a small transpiler of sorts to turn it into

Where is the DevOps for Yocto?

Sat, 27 Sep 2025 00:00:00 +0000

The cozy world of DevOps has spoiled me with tools like Ansible, Terraform, and Packer. In the situations where I have a Linux machine already booted up, and sometimes already connected to the Internet, I can make things really easy on myself by slinging these tools in my usual way for repeatable, reproducible machines that may or may not come with the attendant headaches of immutable infrastructure (sorry, Nix, I’m not in love with you, I’m in love with the idea of you).

It's okay to solve a problem twice

Fri, 22 Aug 2025 00:00:00 +0000

Quoth “How to Become a Hacker”:

2. No problem should ever have to be solved twice.
Creative brains are a valuable, limited resource. They shouldn't be wasted on re-inventing the wheel when there are so many fascinating new problems waiting out there.

Much more at the original post, including preemptive rebuttals to what I’m about to describe. This post is entirely about cataloguing my own error.

I took this maxim, perhaps, a little too close to heart when I was starting out as a software guy, 15-odd years ago. This is my apology to past me. Past me, I apologize.

You don't need CGO to use SQLite in your Go binary

Fri, 18 Jul 2025 00:00:00 +0000

At least not for most use cases. You can just use modernc.org/sqlite instead as your SQLite driver.

For people who aren’t in the Go know, “pure” Go programs are trivially easy to compile cross-platform to all the major platforms by default. You read that right - you can just go build a single Windows executable, Mac executable, and Linux executable on the same machine and just ship it:

# This can all happen on the same box!
export CGO_ENABLED=0 # no c cross-compilation please

export GOOS=linux
GOARCH=amd64 go build -o hello-linux-amd64 hello.go
GOARCH=arm64 go build -o hello-linux-arm64 hello.go

export GOOS=darwin   # aka mac
GOARCH=amd64 go build -o hello-darwin-amd64 hello.go
GOARCH=arm64 go build -o hello-darwin-arm64 hello.go

export GOOS=windows
GOARCH=amd64 go build -o hello-windows-amd64.exe hello.go
GOARCH=arm64 go build -o hello-windows-arm64.exe hello.go

This was the real reason I chose Go over Python for tsk, my instant-search Finnish to English pocket dictionary. I wanted to be able to give Windows users a single .exe they could just run and have work out of the box.

LLM, JavaScript, GitHub Pages, localStorage: A recipe for free apps anyone can use

Wed, 18 Jun 2025 00:00:00 +0000

Earlier today on Hacker News Scrappy made the rounds, with the explicit tagline “make little apps for you and your friends”. I always like to see new projects in this vein.

That’s why I’d like to outline my alternative approach, which

Works cross-platform and on mobile devices by default,
Doesn’t require any app store tomfoolery,
Has great uptime built in,
Gives you just enough data persistence to not get in your way, and
Is owned by you, forever.

This is the stack I used to build my diet checklist, which I keep as a little icon on my phone’s home page. Here it is:

Cross-platform TUIs are easier than cross-platform GUIs

Tue, 20 May 2025 00:00:00 +0000

Below is a GIF of tsk, my pocket Finnish-to-English dictionary, running in my terminal emulator of choice under Linux. It’s what the kids call a TUI, a graphical program that just happens to drive its graphics using terminal graphics instead of graphics-graphics.

Insert GIF here.

You can probably tell that this program fits neatly into the “home-cooked meal” clade of programs. There is a very straightforward problem I want solved - fast, single-executable-portable dictionary lookup, with a few conveniences for the busy language learner layered on top. I am quite happy with tsk in its current iteration and don’t plan to add much more to it anytime soon. It would still save me a lot of time and hassle if I were the only person who could use it.

Binary search isn't about search III. Loop invariant of rightmost element search

Sun, 09 Mar 2025 00:00:00 +0000

If you followed “Binary search isn’t about search” and “Binary search isn’t about search II” properly, the following statements should suffice as a summary:

1
2

L[0:l] < T < L[r:len(L)]    # ordinary binary search loop invariant
L[0:l] < T <= L[r:len(L)]   # leftwise binary search loop invariant

Let’s complete the triptych. Take a wild guess what invariant we use for rightmost element binary search:

`1`	`L[0:l] <= T < L[r:len(L)] # rightwise binary search loop invariant`

A motivating example

Consider again an array like

Binary search isn't about search

Fri, 07 Mar 2025 00:00:00 +0000

Suppose you’re trying to track down a bug that appeared in a series of Git commits. You’ve been idly keeping track of where this bug appears in your lucky commits by hand, while busy with other things. So far you’ve compiled this table:

0000000    🧼 clean, no bug.
0000001    🧼
0000002    🧼
0000003    🧼
0000004    🧼
0000005    🧼
0000006    🐛 bug first appears here.
0000007    🐛
0000008    🐛
0000009    🤔 bug mysteriously disappears...

Then an EMP explodes in your vicinity and scrambles your memory circuits. The Internet is down, you don’t have the git-bisect man pages downloaded locally, and all you remember is that your first commit was good, your last commit was good but for reasons you don’t understand, and something bad is probably still lurking in there. But where?

SQLite is learnable

Tue, 18 Feb 2025 00:00:00 +0000

This is a response to pid1.call’s “Siren Call of SQlite on the Server”, which itself is a response to articles like Wesley Aptekar-Cassels’s “Consider SQLite” espousing SQLite as a server-side technology. Cards on the table, I both love SQLite and think pid1 has the more correct take here.

When I decided on a dime after college to move countries and be with my wife, part of the package deal was that I had to throw away my dreams of easing into the software industry by resting on the laurels of my strong, but not MIT-level-known-worldwide-strong, alma mater (sorry Wildcats). Electrical engineering was just not going to be feasible for a then-monolingual English speaker in Finland, and besides, I majored in it 90% out of curiosity anyway. I always intended to return to my once and future home, the shell, after my Rumspringa with electrons.

The self-hosted to DevOps engineer pipeline

Wed, 18 Sep 2024 00:00:00 +0000

What’s the best way to get a job? Show someone with a job to do that you can do the job within their iron triangle. What’s the best way you can show someone you can handle a complicated k8s deployment, with 7 different CNCF-approved add-ons, zero-downtime rollouts and a whole bunch of YAML files? Probably by competently and publicly running your own complicated k8s infrastructure.

Self-hosters remind me a lot of the sysadmins of yore, who mostly ended up in the profession because they just couldn’t help but mess around with their underlying computing machine until they knew all kinds of weird nooks and crannies within it. I trace my own lineage in software deveopment back to the day my parents finally purchased a Dell laptop and a 300 Kbps Internet connection (residential wiring in Boston left something to be desired), and promptly broke the Windows registry and installed Ubuntu without them ever realizing anything had changed. The next year I got my first internship through a high school program as a Unix admin intern at Akamai, and the rest is history.

How I ask GPT-4 to make tiny Python scripts in practice

Thu, 22 Aug 2024 00:00:00 +0000

First get a working script. “Hey GPT-4, write me a ChatGPT script that does .”
Manually check over the script and iterate until it’s giving me what I want.
“Now wrap the script into a click command-line interface.” I almost always specify to use an --input flag and an --output flag.
If the data it’s working with is human-readable, “Make it so that if --input is not specified, it reads data from stdin.”
If the data it’s writing is human-readable, “Make it so that if --output is not specified, it emits data to stdout.”
If the script is complicated enough that error logging is warranted: “Add logging and colorlog. Ensure all error messages are written to stderr so that it does not clash with output.”

This advice is optimized for small Python scripts, usually under 200 lines in total, for automating semi-mundane tasks like

My pet theory of how great software gets started

Tue, 16 Jul 2024 00:00:00 +0000

(Inspired by yungporko’s Ask HN post, which got me thinking.)

Pretty much every community, dojo, workplace, subculture, scene you can imagine in the modern day had a software sub-scene embedded within it. It can be as small as “that guy who does our Excel”, or as large as the scene itself . This is owing to the fantastic generality of software as a way to make almost anything more efficient, but we won’t go on that tangent now.

Valuable software is about letting people do new things

Fri, 21 Jun 2024 00:00:00 +0000

Today I released a dump of six months of flashcards I autogenerated from my tiny Finnish news archive to make the lives of my fellow language learners easier. The actual code which generates this archive is about 300 lines of Python . The basic value add for the user:

What you want: Better fluency in Finnish.
What you need: Practice. Lots of it.
What this gives you: GOTO 2.

This is emphatically not the kind of product, or use case, I would stumble upon as a brain-in-the-vat developer. It is a niche offering, one I only found because I actually got out into the world and tried to do something hard. In so doing, I recognized the possibility of building a tool that could help me in this non-programming endeavour. So I built it. And I’m proud of it.

Layers of abstraction for me, not for thee

Tue, 18 Jun 2024 00:00:00 +0000

Consider the problem of “how do I run more than 1 terminal at a time”. At this moment, I have at least 5 different ways I can effectively solve this issue:

I can, from a different physical computer, SSH in to a new session.
I can, from the same physical computer, switch to a different tty session with … C-S-F2 through 6 or something. (Rare, but sometimes it comes in really handy.)
I can, in the same user session, open 2 separate windows of a terminal emulator.
I can, in the same terminal emulator, open a new tab.
I can, in the same emulated terminal, run tmux and open a new pane.

There are yet more exotic options like Serial over LAN that I’m only theoretically aware of.

OpenBSD, the computer appliance maker's secret weapon

Wed, 05 Jun 2024 00:00:00 +0000

Between our ESP32 prokaryotic organisms and our 24/7 Internet-enabled megafauna servers, there exists a vast and loosely-defined ecosystem of things the B2B world likes to call computer appliances. Picture a bespoke Pi 4 packaged up neatly with some Python scripts, a little fancy plastic embossing, and maybe a well-guarded id_ed25519.pub in case you end up in hot water during the (long - very long, stable cash flow for generations long) maintenance contract, and you’re in the ballpark.

Doing is normally distributed, learning is log-normal

Tue, 28 May 2024 00:00:00 +0000

There are few things I think about more than the essays on gwern.net, and there are few with as satisfying a theoretical payout to contemplate in my orb as his essay on “leaky pipelines”, aka log-normal distributions.

The skulk: Say you’re working on a Laravel web app. You’re about 90% sure you know how to start the app. You’re 80% sure you know how to handle the infra you’ll need to get it online. And you’re 70% sure you know how to get your first customer. What is your chance of successfully going from zero to first customer? 0.9 * 0.8 * 0.7 = a little over 0.5. That’s … a lot less encouraging than any of the previous numbers, if you buy my multi-step modelling.