However, this year a Ford SuperVan 4.2 made the Nordschleife in 6:48.393, so even without Sabine Schmitz a van was faster than the BYD.
Funny you mention the Ford SuperVan because that’s much closer to the 919 Evo in the "no homologation no limits" category than anything you could register and drive off a lot. A fairer and much more impressive benchmark is the road-legal Ford Mustang GTD running a 6:52. That's still far quicker than the BYD, with roughly two thousand less horsepower.
There’s usually a very real and very hard to describe data related impracticality that voids the usefulness of a design that appears well thought out and complete.
Additionally enterprise AI products are built on custom integrations, and complexity of maintenance overwhelms the engineering team and leaves very little time to build out new things.
The simplest changes that come from knowing insider customer experience have significant impacts. If the default range for a duration filter is 5-30min, and it turns out the most interesting data is really on 1.5hr+ rows. Or adding search across legacy platforms that bury uniform information under deeply nested modals, which people spend 20+ a week clicking through to collect a usable sample set based on existence of a few keywords. But building a system that returns good search results is the hard part.
I do like the “build on top” pieces in your gallery. If it’s fast and reliable enough to collab during a discovery meeting, or a customer success meeting, that would be genius. Because then you’d have a way to pull customers into the right mindset to articulate frustrations with their current software, iterate on getting those frustrations get translated into concrete designs together, and at the end you walk away with something that proves you both understand and can solve their problem to any audience.