Software Should Work 2026
Posted on
This week I had the pleasure of attending Software Should Work (2026), henceforth SSW. This was my first time attending a conference and it was really fun so I thought I would write about it.
Firstly, a huge thank you to the organizer, Isaac Van Doren, for making it happen.
I'll be going through the days and their talks and anything interesting that happened outside of talks. I won't be trying to summarize the talks but what I took away from them and will be handing out made out demerit points for things I didn't like just because I feel like it.
I am not sure how common place it is at other conferences, but diversity was REALLY low. Easily >90% of attendees were white men. You could probably count all the non white men on your fingers. Diversity is not that low at my day job so its a bit confusing to me. If you're a non white man and deciding if you should go to a conference, please do! It really helps make other people like yourself feel included.
# First day
The conference started off on an interesting foot, where protestors with signs "stop gate keeping programming". Cops arrived and the protestors were told they are allowed to protest outside but not in the lobby.
After I came back for reception, I thought I should at least try socializing and ran into none other than Richard Hipp. Turns out he is a regular human like the rest of us and quite friendly too.
# Don't Take the Black Pill - Andrew Kelley
The Black pill refers to something from incel culture, which essentially means give up, its never going to get better.
Andrew tried to convince us, don't give up, try and improve things. There were (what I assume) hand drawn pictures which were a nice touch.
What I took away from the talk is we really need a software dev union. Game devs have already figured it out, but alas I could make a whole other post on software developers attitudes to unions so will leave it out from here.
No obvious LLM usage and I did not feel advertised to so I will award this talk with 0 demerit points.
# Reliability Lessons From SQLite - Richard Hipp
In this Richard went over the history of testing in sqlite and how he and his team has achieved the stability and reliability sqlite has. Hint, it's testing.
Interesting things I took away:
- a quote: "without data you're either lucky or you're wrong".
- DO-178B, a standard for writing critically safe software.
- Design for testability from the get go, test code should be 10-15x the size of your "real" code.
- Test your tests (do your tests actually test things you think they are testing, look into something called mutation testing).
- Richard has beef with golang?
There were pictures/comics on two slides which looked they were AI generated, so unfortunately 1 demerit point from me.
Great talk though, and (for me) the most insightful.
# The Unreasonable Effectiveness of Constructive Data Modeling - Alexis King
A super fun talk. I like types.
- Don't go nuts with your types (do you really need an
Emailtype that is distinct from a string?) - Use your types to move error handling up closer to the caller and let the caller make the decision
- Types + exhaustive pattern matching work best when you use it as a "remind me to not forget to handle a new case I might add later"
No ads or llm usage that I could discern so 0 demerit points.
# How to Find Bugs in Systems That Don't Exist - Hillel Wayne
This was an intro to formal verification and how to model systems. Fun talk and equally fun presenter.
He did advertise his new book (congratulations for writing and publishing!) but did offer a discount code, so will discount my demerit point to 0.5
# Derivations to Deployments: Practical Nix in Production - John Murray
Let's do this off the bat first, 0 demerit points.
John talked about how he uses nix internally at his company to help reduce dev headaches around deployments/CI/local development. Very passionate about nix. Surprisingly (or not?) does not use flakes (in the company).
I approached him after the talk to see if he had any tips or tricks with running a large old ruby monolith through nix. The problems here are that it is unfeasible to run all tests on a developer laptop, and when the tests do run in CI they are split across many build workers so it can't just sit in a singular nix check. Plus when devs are developing code locally they should be only running tests that they are adding or modifying. Unfortunately there was no solution or wisdom shared.
Shopify bought up garnix so that avenue has also dried up. But boy do I wish I could use nix at work.
# Put State in the Right Place, or: How I Learned to Stop Worrying and Love the Web - Delaney Gillilan
Delaney built datastar because frontend development is terrible and javascript sucks and npm sucks (my words).
It's really cute how they have a discord which has turned into backend devs in different languages giving help to each other on how to model things since datastar solves all/most of the frontend js headaches.
0 demerit points because somehow didn't feel advertise-y. It was more of a "here's a thing I built and why" than a "please use my thing" y'know?
# Lightning talks
They were all great, thank you for everyone who presented, and please do more talks in the future!
# The brewery social
I am putting this here as it provides context for the 2nd day.
I generally tried to avoid speakers in general, as they probably enough fans trying to kiss up to them and it probably gets annoying after a certain point, plus what can I even talk to them about?
At this point I was also tired of talking about programming and tech things.
I somehow ran into Filip Pizlo and having a fun conversation around ethics, capitalism and the decline of the USA. On the way back to the hotel he asked if I was going to his talk and if I was going to ask a question.
Well obviously I was going to his talk, I am huge fan of C and Fil-C is dope af. The problem is what question could I possible even ask?
I spent the time before bed reading up on how Fil-C worked so I could ask a "good" question and not a "boring/general" question. More on this later.
# Second day
I ended up being late, guess I had too many drinks.
# Dependency Cultures - Richard Feldman
Again in my own words, javascript's culture around dependencies sucks.
Nothing here that I didn't already know, avoid taking dependencies when you can as they introduce risk and instability to your code.
Couldn't help but feel it was a little advertise-y so 1 demerit point. There was also mention about using claude to help port roc's tokenizer to js to avoid having a syntax highlighting dependency on the languages site, but it was more of a "in passing" comment than a part of the talk so feels unfair to take another point for that.
# System Daydreams and Logarithmic Colored Glasses - Marianne Bellotti
Marianne talked about how we, as humans, are bad visualizing numbers, and it's ok to spend more time thinking about the problem rather than jumping straight into coding. Scale is hard to imagine for us.
This was a fun talk with lots of audience engagement. Marianne is a terrific speaker (to be fair everyone at this conference was).
If I had to pick a favourite talk, this would be it.
I ran into her at the water station, thanked her for the talk and (jokingly) asked if she could be my manager. (I am bad at small talk and again see my comment how wtf do I even talk to these speakers about). Supposedly she was using me as one her "spotters" for the talk. I hope I was a good spotter.
0 demerit points.
# Code Worth Writing - Ray Myers
This talk was a weird one for me.
I'll start with the good. Ray is someone who exudes charisma and confidence and it shows in his speaking. As the kids would say, the man has aura.
The talk started off strong about how software is so bad even non programmers know about the terribleness (e.g. bsod being universally known). Towards the middle it turned into a pitch for a language that his friend is writing. The only thing I remember is the logo looking AI generated.
Towards the end it turned into bun/anthropic bashing? but also (to me) did not feel like it answered the question of what "code worth writing" was? Is it code that can be formally verified?
2 demerit points for feeling like an advert + the llm usage
# Fil-C: Garbage In, Memory Safety Out! - Filip Pizlo
This was my most anticipated talk for the day (see section about the brewery social).
Filip went over how Fil-C works and how it achieves memory safety (it's all here: https://fil-c.org/invisicaps)
Fun talk with live examples and lots of humour.
One thing I noticed in both the talk and the docs, is that there was a lot of mention of heap allocated memory.
One thing that C99 introduced was variable length arrays (VLAs), which are stack allocated. This catches a lot of people by surprise and was generally considered a "bad move". It also makes it really hard to pass VLAs as args because they live on the stack and not the heap.
So for my question I asked how Fil-C handles VLAs and Filip was willing to do a live demo. When it compiled and passed it made me really happy when he said "did you just find a bug in my compiler?" (might be misremembering the exact phrase).
Technically I didn't (the stack write was being optimized out) and memory safety is still guaranteed but Filip was kind enough to file an issue for it, https://github.com/pizlonator/fil-c/issues/265, so I count that as a win.
If somehow you (Filip) ever end up reading this, I hope I managed to ask a good question for your talk.
0 demerit points - while there was a degree of "I'm hot shit, look at what I built" there was still pragmatism in Fil-C being a first step towards something even better.
# Saturation: How Your Software Will Fail at Scale - Lorin Hochstein
This was also a fun talk.
As an SRE at largish company it felt like a walk down memory lane of all the different outages that have happened.
Lorin had a collection of fun outages in his slides, which I asked for after the talk. He said he would post it to the SSW slack (and did) I am preserving the message below (should probably figure out how to get the links working in a code block):
Someone asked for the incident write-up links from my talk. Here they are (along with my own blog posts as commentary)
pckt.blog/b/jcalabro/april-2026-outage-post-mortem-219ebg2
surfingcomplexity.blog/2026/04/12/thoughts-on-the-bluesky-public-incident-write-up
slack.engineering/slacks-outage-on-january-4th-2021
surfingcomplexity.blog/2021/02/08/slacks-jan-2021-outage-a-tale-of-saturation
waymo.com/blog/2025/12/autonomously-navigating-the-real-world
surfingcomplexity.blog/2025/12/23/saturation-waymo-edition
blog.cloudflare.com/details-of-the-cloudflare-outage-on-july-2-2019
surfingcomplexity.blog/2019/08/02/contributors-mitigators-risks-cloudflare-2019-07-02-outage
0 demerit points
# AI & The University - Carson Gross
This was the last talk of the conference, and what a talk to end.
The focus was on how do make sure students are still learning (and more importantly, engaged in the subject material) in the age of LLMs. You could summarize with that meme of Cartman from South Park going "how do I reach these kids?".
More seriously I think we are better off for having teachers like him. It felt good seeing someone care so much about helping students succeed.
Teaching/mentors juniors is my favourite part of my job, though I get to do that less now.
1 demerit point if I am being fair, even if the llm usage is for a good cause.
# Special Thanks
While not part of any talks I would like to shout out the following folks in no particular order:
Alex Gittemeier - For being an incredibly kind and friendly person and putting up with my constant bashing of ruby. I would like to link to your personal site(s) but waiting to hear back if you are ok with me doing that.
David Kotval - For the fun talk about IP (copyright) and LLMs
Shawn Verma - For our ongoing email correspondence
Onorio Catenacci - For the games at Hexagon Alley
Trevor Settles - Also for the games at Hexagon Alley
(Unnamed) - Also also for the games at Hexagon Alley, sorry it was late I didn't quite remember your name but I do remember you live 2 hours east of a major city (purposely being vague because I don't want to leak your location).
Filip Pizlo - For the fun talk at the logboat brewery