All the big boy AI stuff runs on servers, which certainly aren't going to be running on Windows or Mac anytime soon. Linux _is_ what all this NVIDIA hardware is going to run on until... what exactly? FreeBSD overtakes Linux in the server space?
> Who knows what i sthe next big thing after AI. Maybe something that doesn't need Linux?
And Nvidia hardware was running on Linux before AI as well...
> And I'm not so sure that Mac is out of the equation
Mac servers exist but anyone serious about running at scale isn't using them for a number of reasons, not least of which is that Apple simply doesn't make real servers anymore...
They have the money they have the capability design and engineering in house and they are vertical. The question is whether or not they have the will, one thing that might help them come to a decision is that they now have an engineer CEO from the Apple Silicon, part of the company and not a bean counter.
I wish both srcset and <source> elements wouldn't be so redundantly verbose by forcing you to specify the full (relative) URL for each image when they usually only differ by a small part specifying the resolution or format. Sure, transfer-encoding will mostly nullify the overhead but it's still annoying when editing the HTML manually.
Ah, don't edit the HTML manually. Name your images brain-dead predictably, and use a script with some regex.
My blog has some code that searches for <img> tags just after markdown conversion, then uses the src to find similarly named images (image×1.jpg, image×2.jpg, etc.) to use in <source> tags.
One more thing thats nice with <picture> and <source> and <img> over <img> with srcset is the former can responsively change images, while some browsers (Chromium based IIRC) will assume srcset corresponds to the “same looking” image.
IIRC Chromium based srcset will never “downsize” to a smaller image, whereas <source>’s media is like a real media query.
MySQL was a proven solution back in the day, i.e late 2000s. Github/Twitter/Heroku etc were using it. In the past 10-15 years, Postgres has come a long way.
MySQL was always behind in terms of features. In early 2000s it had very limited constraints. Most people were running in ISAM backend and did not even have transaction support. It was being used by people who did not understand how advanced relational DBs were being used. What changed is Postgres overtook its actual competition, which were Oracle, SQL Server and Sybase.
MySQL caught up as well as far as I know, but it still may have some poor defaults that are widely used.
In the early days, PostgreSQL was so much more awkward to deal with. Crash-prone at first, and then there was the whole business around having to drop the db during upgrades. It didn't really match MySQL operationally until around 2002.
During the dot com era it was common to develop and launch on MySQL with the intention of migrating to something else if they became successful (though your typical LAMP stack developer regarded Oracle and SQL Server as being deeply 'weird', so many were willing to stick with MySQL despite the well-known limitations of MyISAM).
From where I'm standing, it seems that PostgreSQL became clearly preferable for new projects from the mid 2000s onwards, but it was only the Oracle acquisition that began to push existing users off MySQL.
I would agree with that, although I personally found it preferable in 2001-2002, since it matched the feature set of SQL Server in all the important ways that MySQL didn't. I think it may have been late 2000s before it was clearly preferred for new projects and even later before it was the thing you had to explain why you weren't using it.
edit to add: MySQL did have operational advantages even later than mid 2000s since it had master/master replication from very early (I don't know how sound it was, I doubt it was perfect). That was a real reason to choose it. We still don't have it in Postgres without extensions and even there citus is not really the same.
Oh yeah, we were early adopters of MySQL's master/master replication at an online retailer. Certainly an "interesting" experience, and I remember becoming very well acquainted with the vagaries of the binlog!
(I remember it being heavily touted in the first edition of the O'Reilly "High Performance MySQL" book. The second edition was about twice the length, with most of the additional pagecount going into detailed explanations of why you should actually be very careful and do lots of testing before deciding to rely on master/master!)
Actually, I remember that one of the drivers away from MySQL before Oracle came along was the 4.x and 5.x period, when there were various performance regressions. And even when those were sorted out, the introduction of InnoDB made people realise that MySQL's apparent speed advantage was really just down to MyISAM lacking referential integrity and transactions.
Certainly, there were people clinging on to MySQL 3.23 for read-heavy data warehousing applications for a very long time.
> And even when those were sorted out, the introduction of InnoDB made people realise that MySQL's apparent speed advantage was really just down to MyISAM lacking referential integrity and transactions.
Yes we saw the same thing play out with MongoDB. Ack without fsync is indeed very fast.
PostgreSQL didn't have real Windows support until... 2010? And even for LAMP, a lot of devs at the time were on Windows, they just deployed to prod on Linux.
The only implementation which is going head on with MRI is JRuby. I've heard from JRuby implementers that keeping the runtime compatible with MRI is an uphill battle. I don't know if AI can help this cause.
I still believe in writing good prompts or good instructions. Bad prompts can sometimes blow up the bill. A poorly written spec can waste a lot of tokens
Agreed - I know it's poor security but damn does it work so well
I'm ok with the risk because I typically am pretty explicit about telling the agent what to do - I don't do the loops like "Do this until X" where the agent can make up its own workflow
When i tell it to add features, it doesn't try to do crazy things like installing packages or making up new paradigms - I usually tell it to do those things when I need to
Maybe this is security cope but at this point you'll have to pry unrestricted yolo mode from my cold dead hands. Maybe I'll change my mind when I pwn myself accidentally
I have a tough time with computer security because it's generally inconvenient and results in a worse developer and user experience
reply