Before anyone downvotes this guy, spend some time on the official Apple support forum.
I can't point at a bug that I've seen addressed in subsequent OS releases.
Seizure-inducing HDMI flickering from Night Shift. Finder Trash not supporting put-back _sometimes_. Printers becoming permanently "paused" sporadically, or worse, very consistently. Mouse lag/stuttering because you used "the wrong USB port." Apple photos libraries corrupting themselves with no recovery paths.
It would be strictly better to just not have the forum, then shouting sorrows into the void would feel more solitary.
I drove a Scitex RIP (raster image processor) as my part time job in college. The CMYK (and if you were rich, N more spot colors) rotation degrees were held in religious fervor. Me: "These degrees seem random." Boss: "You are not to ever digress from those, they are the Only True Values" (and if you play with them in TFA you'll see it indeed becomes a moire mess with tiny tweaks, which would cost a ton in reprinting and delays).
It was mildly humorous that stochastic dithering (even with dot-gain compensation) was an option back in 1990, and produced much higher grade color reproductions, but clients wouldn't have any part of it, because it wasn't "the look" that I think they expect their readers expected.
I maintain a fork of sqlite-vec (because there hasn't been activity on the main repo for more than a year): sqlite-vec is great for smaller dimensionality or smaller cardinality datasets, but know that it's brute-force, and query latency scales exactly linearly. You only avoid full table scans if you add filterable columns to your vec0 table and include them in your WHERE clause. There's no probabilistic lookup algorithm in sqlite-vec.
You're absolutely right—sqlite-vec currently only supports brute-force search, and its latency does scale linearly with dataset size. We did some rough comparisons using its benchmark tools: on the SIFT dataset, latency was around 100ms; on GIST, it was closer to 1000ms. In contrast, with zvec's HNSW implementation, we get ~1ms latency on SIFT and ~3ms on GIST, while achieving recall@100 of 99.9% on SIFT and 97.7% on GIST.
FWIW all of these projects rely on ExifTool (which people should donate to!) and my open-source node.js wrapper (that adds concurrency, does a ton of extra parsing work, and makes things a bit more ergonomic to live with): https://github.com/photostructure/exiftool-vendored.js
Gobsmackingly poor deals made by city and town politicians are par for the course, and why https://www.strongtowns.org/ should be prerequisite reading for any council member, mayor, or board member approving deals that impact their community.
It's easy to look at a glossy project 2-pager and only see the immediate tax revenue.
It's much harder to glean a nuanced understanding of future financial burdens from a given project. No company will have any incentive to be forthright with that information.
reply