A break from programming languages

It is rather damning of our industry that doing this professionally feels unrealistically idealistic to me.

It’s plenty realistic if you work on an internal tools or cost center team, especially in a non-tech-sector employer.

Weirdly Clojure has a similar vibe where most job openings with it recently have been in finance.

The same is true of OCaml, where Jane Street dominates development.

Weirdly Clojure has a similar vibe where most job openings with it recently have been in finance.

Finance is profitable enough that you can excuse a smaller hiring pool, and weirder code. A lot more profit means less pressure on engineers to make specific decisions, so long as whatever it is ends up profitable.

For the first time in 10 years I am working on software not related to finance. I can’t begin to describe how much I hate writing fintech SaaS. Infinite business rules with traps everywhere. You might cause problems for forgetting how to use floating points and decimals. I don’t get excited reading IEEE standards on floating point and fixed point arithmetic. Rolling back long transactions is a big pain and there are important business invariants that touch every single part of your data at once. Customers are incredibly impatient so you introduce features faster than you can organize them and you end up if your app coupled to data (IF customer_id…).

And none of it is cool or exciting.

Now I’m on a chat-ish ia app for sales, and it has been so much fun.

Weirdly Clojure has a similar vibe where most job openings with it recently have been in finance. I also feel that it makes it harder to summon any kind of enthusiasm working on tools for Clojure developers compared to my other work.

Not sure how that’s weird? A lot of their business being in finance was known for years. I mean, Cognitect was acquired by Nubank in the end.

This was a really good read. It was not at all what I was expecting – my expectations were set super low by other blog articles over the years with roughly the same subject.

I do think that joy is seriously under-rated as a motivating factor in our industry. I was saddened to read, though, how much the “being the only woman in the room” was de-motivating for the author. So much that we can do better on, in this regard, but it will likely take concentrated effort over decades to balance this better, so “previously” was the best time to start, and “now” is the second best.

I’m a fan of the PL work of Alexis that I encountered – on the Racket side, and then the support for primitive control operators in GHC which I think is really cool work. Best wishes!

A part of the post that resonates with me is that being disconnected from real-world users can be difficult in the long term. Tooling work (including PL design and implementation) is sort of second-order: we build things for people who themselves build things. Sometimes doing too much second-order work makes people feel a bit lost:

• Maybe there are things in the world that we really would like to improve, and we feel that second-order work is maybe not important enough anymore.
• If we evaluate our first-order impact through the things that our users have built, we have little control over that and it may not be to our liking. (Structurally, the industries that use our niche tools, especially the correctness-oriented ones, are industries that have a shitload of money to spend: finance, blockchain, big pharma, defense.)

(Some mathematicians have the same kind of doubts except worse, as they work at an even higher “order” in that sense.)

I think that moving away from second-order work due to these issues is a perfectly valid thing to do. Here are some other approaches that I think can also work:

1. Do some first-order stuff for a bit, maybe on a break: go write useful software with non-programmer users. Sometimes people manage to share their time between both. Taking a break from second-order work can rekindle motivation and provide important insights on what we aim for.
2. Pay attention to smaller user groups that are less visible and do first-order work we care about. They are probably not the big industrial users, not academics either, and we hear about them less often. Meet them, collaborate, tune our tools in their direction, build a shared sense of community. Maybe it doesn’t matter if you don’t care what hundreds of people in the blockchain industry do, as long as three kind people in their co-op are doing something really nice with your tools.
3. Sometimes we can shift our perspective a bit and think of our users as, you know, also being in the real world, and our tooling work being an act of care. Whenever we improve an error message somewhere, people lives become slightly easier. This matters, but many of us are too busy running to the next issue/PR to take the time to bask in these micro-satisfactions. (I find it easier to think this way about user-interface-oriented stuff than about big PL features.)

Regarding John Backus, I’ve just watched the oral history he delivered to the Computer History Museum, and I get the feeling he wasn’t very interested in programming as a craft. In the 50s, if you did anything with computers, there was no way around programming. This article hints at how engineering is distinct from research, and to me, ‘craft’ is opposed to engineering. If it’s opposed, why pick Backus to represent the craft? Alexis in the paragraph immediately after the Backus quote:

We enjoy programming so much that we are willing to spend enormous time, thought, and effort working on programming systems precisely to free ourselves from the tortuous burden of writing programs.

Technology for technology’s sake doesn’t seem to represent what Backus feels, for example at 1h24m33s, he had such a bleak outlook for the whole field:

Booch: What do you think is gonna happen in my lifetime?

Backus: I don’t know, but I don’t envy you, I am afraid. We’re getting more and more technological, less and less human oriented. As a country we’re getting tremendously aggressive, and we’re gonna pay for it.

A bit later:

Booch: […] Any advice you might offer, if you were to talk to someone from high school?

Backus: Don’t go into software. It’s just such a complicated mess. Frazzle your brains before you try to do anything.

Alexis mentions how she’d spend a dozen hours a day in solitude, working on a problem, but no longer wants to do that. I wonder if Backus had a similar kind of development, since he remembers people so fondly, but never has an emotional attachment to any piece of technology, e.g. at 29m5s:

We had a great deal of fun. Very nice group of people. My main job was to break up chess games at lunch time.

He even mentions how he had no technical interest after retiring at 65 years of age, 1h9m40s:

Booch: What did you do after retirement from IBM?

Backus: I didn’t do much of anything, actually. […] I did a lot of work helping my wife, get her computer stuff.

Booch: So you were tech support for the household. And that was a time frame where there was such an explosion of activity in Silicon Valley. Do you have any impressions of that time frame?

Backus: No, cause I wasn’t involved. I had sort of gotten out of technical stuff, I was more interested in music.

Thanks for this. I very much relate to your tension between inherently loving programming itself, while simultaneously finding the most meaning in work that’s anchored to reality - “real” programmers writing “real” software, whatever that means. Or generally, just the shift in mentality from “I want to write this because I think it’d be cool!” to actually thinking hard about the meaningful way that I, specifically, can contribute to the world’s problems/needs/wants. Somewhere around three years ago (I’m 28 as well) I decided I’m more-or-less done writing code just for myself, and it just feels right.

The urge to strive towards the impossible One True Language is also super relatable. Although my “PL experience” is mostly in the form of being a language polyglot (with a little stint in compilers in college), I was just recently struck with some PL ideas that I’m delusional enough to consider might actually be good and broadly desirable, but at the same time I’m acutely aware that fully realizing them would take many years of insane dedication, to almost certainly be followed by nobody giving a shit. I deeply feel the need for my intellectual drives to nevertheless be tethered to something in reality.

Sometimes, I even worry that my PL interests are mostly just a distraction from other “better” interests - is code itself really all that important compared to what you can make with it? Do I like it simply because it’s a common denominator of what’s been in front of me for so long?

Sorry for the ramble - I have no fucking idea what I’m doing with my life.