Bi-weekly Thursday posts are primarily for paid subscribers, and covers things that are bouncing around in Randy’s head, mostly semi-baked. This week is more of a mishmash of items that are taking up space in my brain.
Silly Twitter design change actually happens
Wow, it actually happened. A month ago I wrote about how practically all metrics were conditional and can’t be understood in a vacuum. It was inspired because rumors of a ridiculous change to Twitter’s design where links to articles won’t have the piece of text that shows the title/url and only show the article picture. Apparently it’s an effort to save screen space, but most people hated the notion since it ruins a lot of the usability. Verge reported that the feature actually launched for iOS. I guess we really are going to see how fast they’ll notice issues and walk it back.
I really do miss the interaction with friends on Data Twitter, but it’s too annoying for me to go back. Just barely getting by with an unearthly mix of Bluesky+Mastodon.
Conversations about all-in-one data software
I have no idea why, but this week’s post about all-in-one software has had more comments than anything I’ve ever written. It was a lot of fun. It had been a struggle writing that post, partly because of sleep deprivation, partly because it’s really murky and I was writing in an effort to think things through.
One thing that came out of the conversations was how the capitalistic chase of revenue growth plays a role in why software tends to bloat with features. Companies can’t just leave a product alone for fear of being out-competed by competitors. They have to sell widgets every day while nonprofit projects like open source don’t have the same level of pressure. At best, an open source project competes in a mythical “marketplace of ideas” but earning market share is not really the point.
Also, a few people mentioned that removing features is extremely rare. I think I’ve only worked on a tiny handful of projects involving removing features. Customers very often complain when any feature is removed because there’s usually at least one person who hangs their entire work process on that feature. It’s far easier to just shove unused features into an “advanced” section than remove it.
In the end, I think I’m just very uncomfortable with the fact that the path of least resistance for product development is to keep listening to user pain points and adding convenience features until you stumble, willfully or not, into building an all-in-one tool. It takes a very strong design vision to avoid that fate.
I need to write a talk =O
Last month someone at work suggested I send a talk proposal to an event. It sounded like a good idea at the time so I dreamt up an idea and sent it in. Who would’ve guessed that three years of forcibly coming up with writing topics would come in handy.
What I wasn’t really planning on was to being accepted. So now I gotta write this thing. I’ve only given 5 or 15 minute short talks before, and only to coworkers. A 30min talk to strangers is gonna be a new challenge.
Anyways, the idea is a fun one to explore so when things are set in stone I’ll share more details about when/where this will happen and whether there’ll be recordings.
Go, 6 years after AI took over
After full clearing all of Armored Core 6, I felt the urge to change up my gaming. Since them I’ve started playing a couple games of Go again after a 10 year break. I’m surprised to learn that while I can only read maybe 1-3 moves ahead from lack of practice, I can still play close to my previous rank.
It’s been 6 years since Deepmind’s AlphaGo famously beat one of the strongest professional Go players in the world. And I was wondering what has happened to the game since them. From the looks of it, some things changed, some things stayed the same.
Much like what happened to chess after AI started beating professionals there, AI tools sprung up to help analyze game positions to help humans study and improve. For example, there’s Ai Sensei which will take my crappy 3-kyu game records and mark out where the biggest mistakes are. That was always an issue with self study in the olden days and we had to rely on stronger players giving reviews or just arduous self analysis. I haven’t caught up with the pro scene yet, but I expect that they’ve massively leveraged the tools and styles have shifted correspondingly. There’s going to be a lot of new opening patterns to learn.
Aside from people needing to guard against the possibility that someone might use AI to cheat in a tournament of some sort, life just sorta goes on. People still play the game for enjoyment. It’s still as much of a challenge to learn and master. Most of us are still roughly as bad at it as we were before the AI came. In an age where the word “AI” is way overused and marketed with doom and gloom predictions, it’s nice to see a corner of the world that just… kept going.