"So I looked for attention signals, and then plotted this web history, taking 50 nodes and converting them into 7, it's a very opinionated view..." Yes, and they're the opinions of an old white guy who has been doing UX for proprietary browsers for 30 years, and it shows. I'm racking my brain to see if I have any friends who would want this. This is peak solutions seeking problems on the missed connections section of Craigslist.

Oh, this isn't for users. Apparently. It was apparently research for the history/storytelling about data that people also will not use and will find invasive. Good shit, good shit. Love it up. Keep it up in Q4, we have to keep momentum up for acquisition.

In the Q&A, a member of the audience asks about the security implications of storing this much data, and compares it directly to Windows Recall. The presenter, Scott, says a few right-sounding things, already having some language for this problem: that you make the world's greatest honeypot, that this is why he focused on collecting attention metrics instead of content, and that Microsoft implemented the feature very stupidly (and that WE can do it better). That last part in particular plays well to the ego of the audience.

But Windows Recall wasn't just badly implemented. It was badly conceived. It was data collection that nobody wanted, all risk, no reward. Not seeing that, makes it seem even more likely than it already did that Scott would fall into the same concept-level failures. His point about people storing their bank credentials on computers doesn't invalidate this point, it makes it more harrowing. Linux desktop security is not okay (which is why misguided people reach to containers for salvation).