I worked at Twitter part-time starting June 2007. I've never talked much about this in public. I'm revisiting it because of the complete disaster Elon Musk has made of Twitter. His sabotage of the company has felt personal to me. It hurts to watch him destroy something I helped create. The recent API debacle particularly stings.
Early Twitter was chaotic without enough experienced engineers. I acted as a management advisor. I helped the engineers organize and the executives work better with engineering. I did some good but I've always wished I could have done more. In retrospect, I should have committed more time. I did enjoy a long insider relationship with some of the leadership and was of some help that way.
The most useful concrete thing I did was what we called "Nelson's graphs". I made some simple measurements of performance like tweet delivery times. Post a tweet to one account and see when it shows up on another account's timeline. (Note this graph shows an average of 15 minutes!)
Simple but useful. It was a clear view of whether the site was working and ended a lot of arguments. These days we'd call this basic devops but in 2008 it was still a novel idea. My graphs kept running for several years even after they leaked to the press.
I grieve for Twitter now. I grew to love it over the years and was an enthusiastic user. Musk has ruined Twitter both culturally and technically. I suspect Twitter will survive in some new smaller, crueler form. But I've moved on to Mastodon and that's working for me.
I’ve found a mobile app for weather I finally like enough to be happy about paying for. Windy, best known for its website. The mobile app has extra phone features like notifications and home screen widgets. Also its UI is a little more understandable.
Windy makes a strong first impression with its colorful animation of winds. But wind speed is not that interesting to me. Windy also does an excellent job displaying radar, air quality, current thunderstorms, etc. Even weather station observations and webcams. All displayed beautifully and uniformly; that’s not easy!
But my favorite thing is the forecast view, hidden away in the website but a bit easier to find on mobile. It’s a lot of detail packed into a very small tabular display. I appreciate that it shows the full forecast by hour going out for days. Also the choice of forecast models; their website explains the options. It’s all very nerdy in the way I want. It’s not great at “weather at a glance” but is good for a deeper understanding.
Sadly they have nothing like Dark Sky’s unique microforecasts. Nothing to say “it will rain where you are standing in 7 minutes”. But they do have excellent presentation of large scale traditional forecasts.
I've added an archive calendar to my linkblog, so you can see old posts going all the way back to 2003. The UI is a little minimal but usable and it will work for any search indexers, which is what I most care about. Note old posts will have a grey background because I wasn't classifying link sentiment (new posts are white or black).
The most fun thing here are the images; I generated them for all my old links. Of course that's only possible if the site is up, I generated most of these in July 2022.
I've got a total of 20,700 links going back 19 years; that's an average of about 90 a month. I've been pretty steady the whole time, 50-150 a month with no missing time. A key thing is it's very easy for me to linkblog a new post, I just use the Pinboard browser extension and it's done.
The time sorted archive is not too thrilling. The other choice here would be to use the tags somehow to extract archives by topic, maybe cluster in to categories or (god forbid) do a tag cloud. A project for another time.
My Starlink Internet service has gotten pretty bad; every evening I'm well under 50Mbps and some hours I only get 2Mbps. (Compare 100Mbps+ last year.) I've given up trying to stream 1080p video at night; that's a pretty dismal result for a new Internet service in 2022.
Starlink imposed major restrictions on US customers last month: 1 TB / month data cap and expected download speeds dropped from 50-200Mbps to 20-100Mbps. Details of all that on my secret blog. Note they didn't drop the price, we're still paying $110/month.
Maybe the new caps will help the congestion? I'm sympathetic to their technical problem. They have limited bandwidth and they have to share it somehow. Caps are an awkward solution; most users have no idea how much bandwidth they are using or why and thus can't control it. Starlink's caps are nice in that if you exceed the cap you just get lowered in priority, not charged money or cut off. So maybe it'll be self regulating.
My real fear is that instead of improving service the result of all this is Starlink is just going to add even more customers to an already overloaded network.
Ken and I just got back from a 23 day trip exploring most of Austria. We had a lovely time although we did get a little worn out and ready for home after a couple of weeks. Along the way we stayed in Vienna, Graz, the Wörthersee, Zell am See, Innsbruck, Salzburg, Hallstatt, Linz, and Dürnstein. A nice mix of cities and countryside. I tweeted a bunch of postcards, easily viewed here and here.
My favorite places were the town of Graz and the countryside in the south from the Wörthersee through Zell am See to Innsbruck. The drive over the Grossglockner Alpine Road was a particular highlight and the lush green alpine valleys of Styria, Carinthia, and Tyrol were just a delight. Salzburg was also a very good visit. Linz was an industrial disappointment and Hallstatt was tourist hell; skip it unless you are very interested in the archaeological story.
We ate very well on this trip; Austrian cuisine is more interesting than the schnitzel-and-boiled-beef that was my stereotypical view. Lots of fresh fall ingredients. The pumpkin cream soup was particularly good. Enhanced with Kürbiskernöl, pumpkin seed oil that's deliciously nutty and green tasting. And of course plenty of fantastic sweets; the French call pastry viennoiserie for a reason. The single best meal we ate was at Pfefferschiff in Salzburg.
This is gonna sound silly but one of the nicest home improvements we've done recently is install a new garage door opener, the Liftmaster 87504-267. It works so much better than my old insecure garage door!
Internet access is the surprise best feature; I use it all the time. Mostly to walk in and out of the garage door without my car. There's also a keypad remote I can mount outside so someone can punch in a code to open the door. Setting this all up was easy and reliable. There's even a way to give Amazon access to open your garage for deliveries.
The opener also has a camera. I would have skipped that if I'd known, saved some money. But it's actually quite useful! If I get a notification the garage door opens I can easily see on my phone what's going on. Basic live views seem to be free, there's a subscription if you want stored video.
Ken's favorite feature is the motion sensor that turns the light on. The lights on the unit are bright, it's enough to light the whole garage without having to flip a switch. I also appreciate there's a battery backup built-in so if the power goes out I can still easily open the door. The drive itself is smooth and quiet too, belt drives are a real improvement.
The #1 upgrade you should do for an older house is a dishwasher; they got a lot better about 15 years ago. But #2 may well be the garage door opener.
I launched a new website for my linkblog; go check it out! I'm proud of how it looks.
My linkblog is a collection of links I find interesting. I write it for an audience, a few links a day of general interest. I've been doing this for 19 years now and I think it's one of the best things I publish.
Link blogging is a venerable form that both was a part of the very first blogs and predates blogs with definitive 90s web sites like Cool Site of the Day. It's no longer fashionable although its spirit lives on in every social media share button. The distinguishing factor of a linkblog is the link itself is the focus. Minimal extra content; I average 10 words a link.
The novel thing on my new website is sentiment. I decide whether the link will make readers feel better or worse and color the post white or black. I think it's important to share negative stories but I don't want to overwhelm everyone with doomscrolling. The other fancy thing is a prominent image for every post. They're a combination of website previews from metascraper and screenshots from shot-scraper.
I wish I had more reach! Only 12 people said they read it on Twitter (despite 7400 followers!). I have more readers via RSS; Feedly shows 100, there's a few more on Slacks. Kind of sad but I accept what I'm doing is not mainstream. I don't think this new web view will bring in the masses but it feels like an important part of making my linkblog more of a real thing.
Way back in 1994 I wrote an undergraduate thesis for my math degree at Reed College. It was a fun project, studying a discrete dynamic system that was an extension of the Ising model. Sort of cellular automata meets statistical mechanics.
A few years ago my colleagues from the Santa Fe Institute wrote a preprint from that work, Vortex Dynamics and Entropic Coulomb Forces in Ising and Potts Antiferromagnets and Ice Models. They were kind enough to list me as an author even though I barely understand half the paper! I do have the pretty pictures, though, plus a healthy appreciation of the complexity of discrete systems.
I've never met one of the co-authors, Cosma Shalizi. But thanks to his having a weblog I now know more about him than the other guys.
Google has agreed to settle a gender discrimination suit brought by employees. Congratulations to the plaintiffs, suing your employer is a difficult and stressful thing. But while the $118M headline looks big it works out to only $7600 per employee, less after the lawyers' generous cut. The harm was "the company paid female employees approximately $16,794 less per year". It's hardly restoring equity.
I faced a similar tiny settlement amount in the 2010 antitrust suit when it was found that Google, Adobe, Apple, Intel, Intuit, Pixar, Lucasfilm, and eBay were all colluding to not hire each others' employees to illegally suppress wages. The first settlement "my lawyers" agreed to was so low that the judge threw out the agreement saying they should do better. The final amount was still just a few thousand dollars for each employee.
I was mad enough I wrote a tart letter to the court when I opted out of the settlement.
I am unimpressed with my nominated attorneys. The fact that Judge Koh ruled the original settlement was “below the range of reasonableness” suggests my attorneys are poor negotiators and have not represented the plaintiffs effectively. ... I’m irritated that they cannot even do class members the courtesy of answering email.
As an unnamed class member I had nothing really to do with the suit, so this letter was my only personal involvement. At least when Gilardi & Co saw this letter to the judge they finally answered my email.
Class action lawsuits are the American way of resolving harm to groups of ordinary people. I understand why the lawyers are well paid for them. But they're seldom a good tool for monetary equity. There is value in getting the company to admit to its bad behavior and make changes.