CAST 2021 Recap

When I hit publish on my article CAST 2021 is AFK a few weeks ago I said:

I’m in an airplane for the first time in years on my way to Atlanta, GA for CAST 2021. CAST is both my first conference in-person and the Association for Software Testing’s first since 2019. I’m pretty excited to confer safely at an in person conference AND to see people from the community.


I was so excited (and nervous) to be at an in-person conference I recorded video of the trip and made a short 2 minute overview of the conference and the AST’s board meeting afterward.

A 2 minute narration of the conference

CAST2021 was meant to be a test of the communities readiness to attend in-person events. As such it was made to be small and safe.

Small meaning no more than 50 people. Safe meaning open only to fully vaccinated individuals (with proof checked upon entry) and designed to be partially open (hence the baseball park). We ended up with 40 excited and energized people learning and conferring over 2 days in Atlanta, GA and it was quite exciting. If that wasn’t enough I probably met 30 of the attendees between conversations, food, a tour of the stadium and of course games, games, games.

Tariq King’s tutorial on Testing AI and Machine Learning

The first day was Tariq King’s tutorial. There’s a ton of information to absorb, we we walked through foundational concepts in AI and ML, while also trying out a number of hands on exercises using publicly available tools such as Teachable Machine, Tensorflow Playground and Google Cloud vision. Tariq used GitBook to list out all of the tools in the tutorial which you can play with here. (Although by now it looks like it has been rebranded from CAST2021).

I’m still digesting what I learned, but one of my biggest questions around AI in software testing tooling space was addressed by Tariq. The primary advantage I’ve seen advertised is around better locators which ideally translates into less maintenance. While useful this also seems rather bland and perhaps not very compelling for my context.

What I learned is the AI/ ML tooling ends up being an abstraction layer around our test automation tools. Rather than having a tool look into the DOM to understand html or javascript elements like many popular tools do, computer vision is used to identify what is on the page. Once the machine knows what’s on the page (after it’s trained) from there you can build a model on what to do. From there it matters less what tool is driving the actual interaction.

Day 2 – Conference Talks

Day 2 of CAST offered the full conference experience with 6 speakers. CAST is famous for it’s use of k-cards and facilitated discussions. I’m always amazed at how a few small questions can lead to tangents and additional clarity during an interactive discussion. I get way more food for thought after this exercise than I do online or in a slack channel (most of the time).

James Thomas (a fellow AST board member and friend) did a great job writing up each of the talks at CAST. I highly recommend checking out each of his articles:

Per usual, James did a great job coming up with funny (pun-worthy) names. He also did some sketch noting which you can find in his articles and via Twitter using #CAST2021.

Ultimately I’m super happy to have gone to CAST. I met a lot of folks from the Atlanta testing community and I learned quite a bit from the talks and tutorials. Two weeks have passed since the conference and I still feel energized from the learning and bonding opportunities I was given. Hopefully I’ll see those same people and a few more at CAST 2022!???

If you enjoyed reading this article or gained some insight from it, please consider following me on twitter, sharing it or buying me coffee!

Subscribe to Shattered Illusion by Chris Kenst

Sign up now to get access to the library of members-only issues.
Jamie Larson