I’m staring at this comment for a while now: With 3ms latency combined per token, wouldn’t that mean (1 / latency) = 333 token/s for the theoretical upper bound? I’m not trying to nitpick, just curious if I misunderstand something.
I have several years worth of timestamped sensor data in a SQlite DB that is 12+GB and growing. One idea was to experiment with DuckDB. Your comment about DuckDB has me worried as time ranged queries are common. Any link for me to dig deeper?
How did you arrive at 82% of salary being taken by taxes and social security? I read the Wikipedia article but I don’t how the numbers would add up to 82%.