If you’re targeting data for people to use on the same call then requires more intense work, while if you’re targeting data for people to use after or on an ongoing basis (eg an established business meeting with staff/vendors/etc) then having a predefined one seems good. Same interface as a CRM: picture, some audio clips for users to choose from, and confidence scores on each section of the recording for cases where people sound similar or are talking over each other. Over time the diarization gets better as users accept/reject/tag samples and that work helps them feel more aligned with the tool.

That’s not “the same amount of people were shown the video and 2x the people clicked in”. The algo is always skewing.

“For a low $/GB, we’ll give you everything from this site and 1000 others as (structured data/a database)!”

(yes there are lots of good counter arguments to this, but before you reply think ahead a couple extra steps)

Specific traces for specific inferencing will mean that some generations get deprecated. Look at H265.

No, no high-traffic sites actually uses P2P hosting but that doesn’t mean we’re not at a point where they can. In fact, I’d say the biggest impediment to P2P sites in 2026 is the absolute unit that is advertiser data. We have P2P databases, and round trip on the earth is ~300ms (aka “quick propagation” can be seconds which is more that enough for 98%). We have VPS and serverless. We also have content addressing and symmetric cert verification, but when a site’s “traffic” is P2P you lose visibility on most of it. And can you imagine the nonsense of advertisers adding client side tracking data to the public db so that they can query it? Woof.

Online, the most money being made is by people who get attention. The making is not just secondary, in many cases it’s a calculation. Even genres that I enjoy as a maker for having real tactile value (repair, restoration, rug cleaning, etc) are inherently not profitable unless they are getting paid out by the video platforms.

For artists this has created a wild set of conflicting constraints. Some can do the Patreon model but for those who don’t, they need attention in order to sell art in order to generate enough money to make art. In this world, AI lowering the barrier to entry means that we’re close to (if not already there) “artists” (aka people farming for money) soaking up all the attention and leaving none for the people who make things.

It’s not that all artists will use AI to pretend to make stuff, it’s that all the pieces are there for a world where these money farmers will be the only visible type of artist.

Now, if they had said “We need 200k to send CC photographers to the Olympics”, count me in. That sounds bigger and way more interesting!

Elasticsearch has been a plague for users who already know what they are looking for.

I’m shocked to see that LORD can’t be used because of licensing. Legally I know that’s how it is, but it Just Feels Wrong(tm) that a popular game based on 9600baud phone modems is encumbered in the era of fibre, wifi, and smartphones