ACCU On Sea 2026 - My Talks

/W4 /WX: How hard could it be?

This was the second time I gave this talk. You can read my reflections on my first outing at BeCpp 2026 here. As a guesstimate, there were probably about 50 people that attended. Maybe a little less. Other than the Shading Languages Symposium (which only had about 60 attendees), I think this would be my smallest audience that I have given a talk to.

This time, I had a 90 minute slot so I put a lot more content into the talk, and during practice sessions I did tend to take between 85 and 90 minutes to get through it all. However despite this, I managed to get through everything in 67 minutes. This was a tad deflating because I clearly went through it too fast despite my best efforts to not do that! There are also some parts of my talk which need to be corrected:

1. I erroneously stated that Link Time Optimization (LTO) was the mechanism for removing dead or redundant code. But it ain’t. Guy Davidson helpfully called that out, and said that it was COMDAT folding (OPT:ICF) that was doing that which is entirely correct.
2. I have a typo on one of my slides. This was quite ironic considering that entire slide was demonstrating code that had a bug in it which was due to a typo! This is funny, and part of me wants to keep it in… but it will be fixed for next time :)

The good news though is that I think this is all the negatives I have about my talk. I feel like everything else was much more positive. I got through everything I did want to say. I can’t really think of me missing anything out, or tripping over my explanations or words massively. I guess I just spoke much faster than I normally do while up on stage. I also did a bit more audience participation this time around. Just asking people to raise their hands. No shouting out of answers or anything like that. And I got a decent amount of participation from that which was fun! I even feel somewhat good about getting Guy Davidson to break one of his rules twice during my talk. During Guy’s morning addresses he always mentioned that people should not shout out in the middle of a talk unless prompted. Save questions for the end. Guy broke this rule on 2 occasions:

1. To try and tell me about the LTO mistake. I made him wait for the Q&A on this one
2. “Why do you have cannons everywhere?”. I indulged this one because it was fun. We make THE pirate game, Sea of Thieves, so of course we have cannons littered everywhere :)

I also got a decent amount of Q&A which felt like it took another 10 minutes, so I made decent use out of my 90 minute slot if you factor in the Q&A. Many of the questions were around extra tooling that we could deploy as incremental steps to fixing some of these warnings. E.g. Taking note of all the warnings that we have and having a CI job fail if that list gets larger. There were some other interesting ideas and tools mentioned too. However, I am somewhat against this kind of mentality for warnings for a couple of reasons:

1. I personally just want these warnings to not exist in the code base. Full stop. That is the end goal. Tooling would imply that some warning emission sites are fine. They are not.
2. Extra tooling involves more stuff to maintain and set up. This gets in the way of just eradicating the warnings.
3. These tools typically require full builds to work. So the best place for them would be our CI. I really do not want people hating our CI because it fails on a warning. Any tooling would need to work with iterative builds.

I see the value in these types of tools when you are working with more fine grained warnings. But we are absolutely not there yet with Sea of Thieves. We are still taking the first steps. The warnings that I am talking about, are warnings that people would take for granted as always being enabled.

Walter Brown

I have to include a little section about Walter here because he is a large reason why I had such an amazing time at ACCU on Sea. First off, it was an absolute pleasure to meet him in person after having watched his various talks for the last 12 years. It was also an honor to have him sitting front row in my talk. We had some great discussion throughout the conference. Some of that discussion was around him giving me some amazing feedback on my talk that I gave at this conference and also on my previous talks. Quite the humbling experience! Other parts of discussion were around the history of software engineering and C++. All very interesting stuff!

Lightning Talk - An Impactful Conversation

I gave a lightning talk in addition to my main talk. I love the lightning talks at ACCU. Always a fantastic time, and you never know what you might get. For example, Will Buchanan, who was one of the volunteers and someone I chatted to a fair amount, got up on stage and proceeded to sing an original song. Once he finished singing, he left the stage without saying a word. It was great!

My lightning talk this year was inspired by a time when I upgraded my compiler and found that I was getting many more warnings being emitted than before. I was able to track down the exact point that those warnings were introduced to MSVC in a reddit thread. I found this exceptionally funny at the time, but also pretty awesome. It helped that it was STL who implemented these changes who is a fellow Microsoft engineer (yes Microsoft owns Rare), and it was connect (performance review) season at the time of the conference.

I had fun giving this lightning talk. I got some laughs from the audience too which I always appreciate.

What I did not appreciate was that I somehow screwed up the start! When I was handed the clicker I must have held down the forward button too long and ended up in the next person’s slides. It took me a few seconds to realize what had happened and to get back to my starting slide and actually start the talk. I think I played this off well enough with some humour, but it was genuinely terrifying at the time! I have no idea what the folks on the youtube channel will do with this one… Will they keep the false start in? Or will they just edit that bit out? No idea. A part of me hopes that they keep it in (it was funny looking back on it), but I am not entirely sure that would lead to a good YouTube video!