Show newer

Andres Freund will join us for Talking Postgres E31 to discuss "What went wrong (and what went right) with AIO" (Asynchronous I/O). Join live on September 17, 2025 @ 10:00am PT in #talkingpostgres on the Microsoft Open Source Discord.
#Open Source

RE: https://www.threads.com/@aaronw.dev/post/DN89pYcEsZA

I cannot state this clearly enough: David Heinemeier Hanson can go fuck himself right into the sea.

DHH is the antithesis of everything the Ruby community has always stood for, and I am once again begging for the #Rails team to completely cut ties with this shitstain, via a hard fork if necessary.

I will be the *first* to switch my gemfile over to a fascist-free Rails.

@gerrymcgovern Ah, and since this seems to be making the rounds: it's time for my occasional reminder of one especially unpleasant aspect of the LLM hype storm.

LLMs take advantage of a faulty heuristic many (probably most, but don't have exact stats) people have about human intelligence: namely, someone or something that produces fluent and grammatically coherent text about a wide range of topics, on demand, is "intelligent".

This is a very compelling heuristic and also very wrong.

In the case of LLMs, it's wrong in the way we're all talking about here: it leads people to see thinking or reasoning in the statistical word salad these beasties produce.

But it's even more of a problem when people invert the heuristic and turn it into "someone or something is *not* producing fluent text about a wide range of topics, therefore they are *not* intelligent". This mirror-world version causes enormous social harm to people who for whatever reason (autism, selective mutism, etc.) can't speak the way people expect they would be able to "if they were smart".

Some pictures from the lightning talks round we had after lunch.

Join us for an afternoon with more talks and debates: youtube.com/watch?v=tJYEuIpzch4

Pictures by honkingelephant.com

Kaarel Moppel presenting a Lightning Talk at the PGDay Lowlands Warmup #Meetup at Lunatech in #Rotterdam

#PostgreSQL #Conference

Jan Karremans presenting a Lightning Talk at the PGDay Lowlands Warmup #Meetup at Lunatech in #Rotterdam

#PostgreSQL #Conference

@Edent also in general window functions are great for aggregating these kind of stats.

@Edent be careful with 'UPDATE count = count + 1" patterns. They're unlikely to accurately count as you expect due to concurrency issues. Especially in MVCC systems, it will depend alot on transaction isolation levels.

For this kind of thing, just logging the visits as records and aggregating on analysis / periodically is often the better approach. Removes a lot of the concurrent write headache and give you more flexibility.

Show older
Mastodon

Time for a cuppa... Earl Grey please!