Global Software Architecture Summit (GSAS) was organized for the 4th time this year. It seems to happen in Barcelona each time. A few of my colleagues went there last year and said it was one of the best conferences they attended so far. This encouraged me to buy the “blind ticket” very early and be really excited about getting to go there. However the enthusiasm faded a bit when it was announced that the conference will be all about Artificial Intelligence (AI) this year and my use of it is very limited.
The two best talks for me where the opening talks each day:
What Makes Software Architecture So Intractable? By Neal Ford. Neal Ford is the author of several books on Software Architecture including Fundamentals of Software Architecture (which I’ve read) and Software Architecture: The Hard Parts. The main idea of this talk was that the Software Architect is as good as the trade-off analysis (s)he performs. There were a few tips on how to be better at this e.g.
The ideas of the talk were demonstrated through the common questions like “Should event payload include all the needed data or just the keys required to fetch it?” or “Should common functionality be implemented via shared libraries or services?”. Neal Ford really demonstrated that there is no single answer to these questions and it very much depends on the context you are working in.
Micro-Frontends anti-patterns by Luca Mezzalira. I’ve recently moved to a Site Reliability Role so all things frontend are far from being relevant for me. But I was familiar with this concept from Udi Dahan’s course on “Advanced Distributed System Design” and I liked what this talk added to my basic understanding. The idea of the talk was that people often mix up micro-frontends with simple components. In reality when designing micro-frontends the mindset should be more similar to designing the boundaries for micro services. Each micro-frontend should:
There should also be 4-5 micro-frontends per page and they should be loosely-coupled i.e. communicate via events. It was curious for me how this resembled the typical guidelines backenders get for micro-services.
All the AI talks. These were less relevant for me, so I want to summarize here just some ideas that stood out to me the most. First of all - the types of AI integrations available today:
The types are listed in a pyramid based on how widely adopted/ easy to access they are. RAG requires a pipeline for extracting the data and augmenting the prompts, so less people will use it. AI agents seem to require even more setup. Majority of the conference talks focused on RAG. The leading suggestion was to set up a RAG system over all the internal documentation of your company to make it searchable and more accessible.
There were also a few talks that provided advice on how to make more use of the prompt based AI e.g. what questions to ask if you are trying to get AI to help you to do a trade-offs analysis or solve a problem. The idea I liked the most was to ask AI to set up a “Think tank” to vet your idea. This gets the AI to model a full conversation between some experts, including asking questions and answering them. Getting AI to ask questions in order to provide a more specific solution is a really tricky thing to achieve.
It was also clear that AI is still in the ‘wild west’ phase of technology - a few speakers said it was difficult to prepare the talk in such a way it wouldn’t be outdated when time comes to present it in e.g. a month. But there is also the prediction that AI tools will become a commodity, an everyday tool at work and those who will learn and adopt them faster might get a competitive edge.
If last year Euruko motivated me to go to more Software Engineering conferences, this year GSAS discouraged me again. Traveling to another country for the conference and spending several days listening to the talks is quite the investment, and the return on it was rather low. Ideally, I would like the conference to either give me ideas for new tools or practices I can adopt at work or present some problems that people faced when using a specific approach so I can be warned to not repeat the same mistake. A colleague told me that last year the talks in GSAS were exactly like that - people “from the trenches” were presenting what they tried at their companies and what worked and what didn’t. This is why he found the conference so valuable. This year the talks were more about O’reilly authors presenting their books.
However, I’m not giving up on conferences yet. I have a ticket to GoLab for November and I would like to attend some SRE conference next year.
© 2024. All rights reserved.