Did another pass on the design for this. I don't think I've explored enough for a prod-level implementation, but this covers the bases we need for an experiment:

Here's a video walkthrough of the design to date and the strategic concerns involved.

Moving to signoff because we will break this out into implementation tickets and spikes.

concept mockups for reference.

Leaving notes here as well for anyone following along:

• I love the visual styling of the machine-generated content. I think it's really good at giving that machine-generated visual cue for those that are not reading the disclaimers
• I think the accordion and the location fit the feature really well
• I like the verified/unverified tags. I wonder if the copy here is correct (I think it also depends on our level of trust in the model itself versus trust in the overall content - see comment below on Simple English)
• I like the level of detail on how the summary was generated and the copy in general
• I would like to push back on using data from simple English alongside this feature (more info on that below)
• I wonder if we can look for other locations for presentation on Minerva. Potentially after the first paragraph/after the infobox

I have doubts about using data from simple English/presenting simple English summaries. A couple of reasons:

• This is not friendly to all of our wikis. Only English Wikipedia has simple English. In addition, this would only work for a very small percentage of articles on English Wikipedia.
• Simple English is not a very active project, information on it can be outdated. We want to use our most recent source of information for data, which is Wikipedia. Given this context and the quality of our model, I would assume that the model summary would actually be of higher quality than the Simple English version as it would be based on the work of more editors.
• Simple English does not provide a summary of the entire article, which is what we’re hoping for our model to do at some point in the future.
• As we learn more about summaries, we will learn more about the level of simplification that works best for the majority of our readers. This might not be the level of simplification available on Simple English
• The goal here, in addition to giving readers summaries, is to automate tasks that editors are doing right now, thus eventually eliminating the need for projects such as Simple English and allowing editors to focus on the main projects.
• Communities might not not be really open to showing content from different wikis

Thanks for those notes @ovasileva !

I'll fold these into the next round of designs on this feature along with other comments I've had from the design team in T374636. I think you're ok to sign this one off, as it only represents the first pass on designs.