Before the weekly check-in meeting, my lead mentor had communicated that we would have a contributor evaluation. It was to be a chance to assess what we'd done so far, and our satisfaction with what we had done so far. He asked us to reflect on our answers before the meeting, as it would be grounds for our discussion.
Weekly Check-in
We started the meeting with a bit of reflection. Storm pulled up a Jamboard and we pasted sticky notes against two columns: one for things that have gone well so far, and the other for things that we felt could have gone better. For the former, we touched on our progress so far and iterative speed, and for the latter, we talked about a few knowledge gaps here and there that couldn't be avoided, and the need for a faster RFC review process.
We ended the meeting with some laid-out tasks for the week and an agreement that I would join the next core team meeting.
The Core Team meeting
The whole idea of my being present at the core team meeting was to answer any questions about the RFC (or the processes in it) if anyone had any. There weren't many questions about it, but it felt nice being there again. Most of the discussions in the core team meeting were about the just-concluded Wagtail Space in the Netherlands, as well as the about-to-start Wagtail Space US.
They also discussed the recent statistics of Pull Requests, Issues, the first-time contributors since the Wagtail Space in The Netherlands was done, and whether or not to accord the numbers to the event. More core team members agreed they'd make time to go through my RFC, and they mentioned broadcasting it to the wider Wagtail community since they would ultimately be the ones the changes would affect the most.
We called it a wrap at that point.
Challenges
It was a slow week for me, as I felt a bit ill, so this was a challenge. Then I needed a headstart on resources to go through for the AI part of the project. For the illness, not much apart from rest could be done, but for the resources, my mentors asked me to start by researching scenarios where machine-generated alt texts would probably outperform the ones by humans. They mentioned finding out enough to make a strong case with the core team because as good as the project was on paper, if it offered no real benefits, then it would just be a lot of fancy work in the wrong direction. My starting point was to be the Accessibility team's research. Afterward, the next step for the research would be on the "contract" of what the AI models would be like. We would have to know what they would require from us and what we would pass to them for a smooth flow. This would help us know how to structure the backends when the time for development comes.
What I learned
I'm still neck-deep in the research, but so far, I've come to learn how much better AI is at generating alt text than human beings on average. Often, this is a case of humans not being interested or dedicated, rather than the case of an inability to do it. I'm still delving deeper into the research, but this is my starting point for now, and I hope to share more of my findings as I progress.
I really do hope I get better soon so I can move even faster with my project and commence development for the AI stretch, which gets me so excited.
It was a good week despite the health setback, and I feel like the work I'm doing starts to matter more and more. I hope to have great news to share by next week.
Thank you for reading.
Cheers. 🥂
Top comments (0)