Probably data-analysis/AI type stuff to track users and advertise "better," making the backend more efficient to reduce costs, and adding support for new hardware. A lot of big, very profitable companies also have skunkworks-like projects for exploring new ideas and prototypes, most of which never make it into production.

Tbh most employees at a company this size become risk mitigation more than anything else. Once you've reached a certain level of success, you're looking at what doesn't move the needle as much as what makes it move positively. There could be a feature that is a major QoL improvement, but because in a test segment it performed 1% worse than base then it won't be implemented.

Spotify, I believe, still works in the tribe and guild model that they created.

Chapter = people with the same skill set, squad = a group of people from different chapters focused on a single project, tribe = a group of squads focused on a large business goal, guild = a collective of folks who have a shared interest like Data Privacy.

Suffice to say, Agile is an imperfect tool and as you try to scale it, you need an increasing number of people to support it and make it run. Coders and Designers are likely just a fraction of their head count.

I've worked places that don't have that support structure in place and they've stagnated for years struggling to get the most basic of decisions made. Decisions is what it is about too. Rarely do you get actual leadership from the c-level and especially from a CEO. So you end up with a lot of cooks trying to work out why the broth doesn't taste quite right and lacking confidence to just add a bit of salt.

In my experience it's always new apps that ruin the existing experience.

Well, they have to make new, broken terrible features and then come fix them when people complain by basically putting it back to how it was.